1

Google LLC, formerly Google Inc., is a United States company specialising in internet-related products and services. It is principally known for its search engine, which allows internet users (also referred to as ‘users’ or ‘consumers’, depending on the context) to locate and access websites that match their requirements by means of the browser they are using and hyperlinks. Since 2 October 2015, Google LLC has been a wholly owned subsidiary of Alphabet, Inc., the ultimate parent company of the group (together, ‘Google’).

2

Google’s search engine, accessible at www.google.com or at similar addresses with a country code extension, enables search results to be obtained and displayed on pages appearing on internet users’ screens. Those results are either selected by the search engine according to general criteria and without the websites to which they link paying Google in order to appear (‘general search results’ or ‘generic results’), or selected in accordance with a specialised logic for the particular type of search carried out (‘specialised search results’, also referred to as ‘vertical’ or ‘universal search results’; ‘specialised search results’). Specialised search results may appear without any specific intervention on the part of the internet user alongside general search results on the same page (‘general results page(s)’), or they may appear alone in response to a query entered by the internet user on one of the specialised pages of Google’s search engine or after links appearing in certain areas of Google’s general results pages have been activated. Google has developed various specialised search services, for example for news, local business information and offers, flights or shopping. It is the last category that is at issue in this case.

3

Specialised search services for shopping (‘comparison shopping services’) do not sell products themselves, but compare and select the offers of online sellers offering the product sought. Those sellers may be direct sellers or sales platforms grouping together the offers of numerous sellers from which the product sought can be ordered immediately (eBay, Amazon, PriceMinister or Fnac being among the best known).

4

Like general search results, specialised search results may be what are sometimes referred to as ‘natural’ results, which are not paid for by the websites to which they link, even if they are merchant websites. The order in which those natural results are displayed in the results pages is also independent of payment.

5

Google’s results pages, like those of other search engines, additionally contain results which, on the other hand, are paid for by the websites to which they link. Those results, commonly called ‘ads’, are also related to the internet user’s search and are distinguished from the natural results of a general or specialised search, for example by the word ‘Ad’ or ‘Sponsored’. They appear in specific spaces on the results pages or among the other results. They may take the form of specialised search results and in fact some of Google’s specialised search services are based on a paid inclusion model. The display of those results is linked to payment commitments entered into by advertisers at auctions. In some circumstances, additional selection criteria may be applied. Advertisers pay Google when an internet user clicks on, and thus activates, the hyperlink in their ad, which leads to their own website.

6

Google’s general results pages can include or have included all types of result referred to in paragraphs 2 to 5 above. As is also explained in paragraph 2 above, specialised search results, whether natural results or ads, may also appear alone on a specialised results page in response to a query entered by the internet user on one of the specialised search pages of Google’s search engine or after links in certain areas of Google’s general results pages have been activated.

7

Search engines other than Google’s own offer or have offered general search services and specialised search services, such as Alta Vista, Yahoo, Bing or Qwant. There are also specific search engines for comparison shopping, such as Bestlist, Nextag, IdealPrice, Twenga, Kelkoo or Prix.net.

8

According to Google’s uncontested account, it began providing internet users with a comparison shopping service in 2002, after or at the same time as other search engines such as Alta Vista, Yahoo, AskJeeves or America On Line (AOL). That initiative was in recognition of the fact that the processes that had hitherto been used by search engines did not necessarily return the most relevant results in response to specific searches, such as those relating to news or shopping. Google thus began providing comparison shopping results (‘product results’) from the end of 2002 in the United States, and, approximately two years later, gradually extended that provision to certain countries in Europe. Those results were not the results of its ordinary general search algorithms being applied to information presented in websites – such information being first extracted by a process known as ‘crawling’ by which Google explores web content for the purpose of indexing it, then selected in order to be added to Google’s ‘web index’ and, lastly, sorted by relevance for display in response to the internet user’s query – but the results of specific algorithms being applied to information contained in a database fed by the sellers themselves, called the ‘product index’. These results were first provided through a specialised search page, called Froogle, that was separate from the search engine’s general search page, then, as from 2003 in the United States and 2005 in certain countries in Europe, they were also available from the search engine’s general search page. In the latter case, product results were grouped together on the general results pages in what was called the Product OneBox (‘Product OneBox’), either below or parallel to the advertisements appearing at the top or at the side of the page and above the general search results, as shown in the following annotated illustration, supplied by Google:

9

If internet users used the general search page to enter their query in relation to a product, the responses returned by the search engine included both those produced by the specialised search and those produced by the general search. When internet users clicked on the result link in a Product OneBox, they were taken directly to the appropriate page of the website of the seller offering the product sought, where it could be purchased. Furthermore, a special link in the Product OneBox directed users to a Froogle results page with a wider selection of specialised product results. Google explains that Froogle results never appeared in general search results, however, while the results of other specialised search engines for comparison shopping did.

10

Google states that, as from 2007, it changed the way in which it developed product results.

11

The changes made included Google abandoning the name Froogle in favour of Product Search for its specialised comparison shopping search and results pages.

12

As regards product results displayed from the general search page on the general results pages, first, Google enriched the content of the Product OneBox by adding images. Google has provided the following illustration of the first type of image addition:

13

Google also diversified the possible outcomes of the action of clicking on a result link shown: depending on the circumstances, internet users were, as before, taken directly to the appropriate page of the website of the seller of the product sought, where the product could be purchased, or they were taken to the specialised Product Search results page to view more offers of the same product. Over time, the Product Onebox was renamed the Product Universal (‘Product Universal’) in different countries (for example in 2008 in the United Kingdom and Germany), while at the same time being made more appealing. Google has provided the following annotated illustration of the two variants of a Product Universal:

14

Secondly, Google established a mechanism called Universal Search which, if a shopping search was identified, made it possible to rank, on the general results page, products covered by the Product Onebox, subsequently the Product Universal, against general search results.

15

As regards paid product results appearing on its results pages, in September 2010 Google introduced in Europe an enriched format compared to that of text-only ads (‘text ads’) that had appeared previously. If the advertiser so wished, by clicking on the text, internet users could see, in a larger format than the initial text ad, images of the products searched for and the prices charged by the advertiser. Google has provided an annotated illustration of such a text ad extension:

16

In November 2011, Google began to supplement its text ad extension facility in Europe with the direct display, on its general results pages, of groups of ads from several advertisers, together with images and prices, which it called ‘product listing ads’ or ‘product ads’ (‘product ads’), and which appeared either on the right-hand side or at the top of the results page. By clicking on an ad in the group, internet users were directed to the advertiser’s website. Google has provided the following illustration of a product ad:

17

Google subsequently discontinued the concurrent display, on its general results pages, of grouped natural results for specialised product searches (Product Universal), grouped product ads, text ads, any text ad extensions and general search results, having decided that it was not desirable for that situation to continue. Consequently Google discontinued Product Universals and text ad extensions on its general results pages in Europe in 2013. As a result, only groups of product ads, renamed ‘Shopping Commercial Units’ or ‘Shopping Units’ (‘Shopping Units’), text ads and general search results were subsequently shown on those pages. Google has provided the following annotated illustration of a Shopping Unit, displayed above text ads and a general search result:

18

Accordingly, internet users who clicked on an ad in a Shopping Unit were always directed to the advertiser’s sales website. They would access Google’s specialised search and results page for comparison shopping, containing further ads, from the general results page only if they clicked on a specific link in the Shopping Unit header or on a link accessible from the general navigation menu (‘Shopping’ menu link).

19

Google states that the selection of ads for the Shopping Unit involved not only the auction mechanism referred to in paragraph 5 above, but also similar criteria to those which it applied to generate its natural product results, referred to in paragraph 8 above. It explains, without being contradicted, that the selection could in certain circumstances result in text ads being ranked higher on the general results page than Shopping Units, or vice versa, or could even result in the latter not appearing at all if the number of high-quality ads was insufficient.

20

At the same time as Google removed Product Universals from its general results pages, it also stopped displaying natural product results on its specialised Product Search results page, which had become a page containing ads only, called Google Shopping. Google has provided the following illustration of a Google Shopping page:

21

The present case has resulted from a number of complaints that were lodged with the European Commission, in or after November 2009, by undertakings, associations of undertakings and consumer associations, as well as cases referred to the Commission by national competition authorities (in particular the Bundeskartellamt (Federal Cartel Office, Germany)).

22

On 30 November 2010, the Commission initiated proceedings against Google pursuant to Article 2(1) of Commission Regulation (EC) No 773/2004 of 7 April 2004 relating to the conduct of proceedings by the Commission pursuant to Articles [101] and [102 TFEU] (OJ 2004 L 123, p. 18).

23

On 13 March 2013, the Commission adopted a preliminary assessment under Article 9 of Council Regulation (EC) No 1/2003 of 16 December 2002 on the implementation of the rules on competition laid down in Articles [101] and [102 TFEU] (OJ 2003 L 1, p. 1), with a view to the possible acceptance of commitments by Google that would address the Commission’s concerns. In its preliminary assessment, the Commission considered, in particular, that the favourable treatment, within Google’s general results pages, of links to Google’s own specialised search services as compared to links to competing specialised search services was capable of infringing Article 102 TFEU and Article 54 of the Agreement on the European Economic Area (EEA).

24

Although Google stated that it did not agree with the legal analysis set out in the preliminary assessment and challenged the claim that the practices described by the Commission infringed Article 102 TFEU, it submitted three sets of commitments, the first on 3 April 2013, the second on 21 October 2013, and the third on 31 January 2014.

25

Between 27 May 2014 and 11 August 2014, the Commission sent letters pursuant to Article 7(1) of Regulation No 773/2004 to the complainants who had lodged a complaint before 27 May 2014, informing them that it intended to reject their complaints. The letters outlined the Commission’s preliminary view that the third set of commitments submitted by Google could address the competition concerns identified in the preliminary assessment.

26

Nineteen complainants submitted observations in response to those letters. After analysing those observations, the Commission informed Google on 4 September 2014 that it was not after all in a position to adopt a decision accepting commitments in accordance with Article 9 of Regulation No 1/2003.

27

On 15 April 2015, the Commission reverted to the infringement procedure provided for in Article 7(1) of Regulation No 1/2003 and adopted a statement of objections addressed to Google, in which it reached the provisional conclusion that the practices at issue constituted an abuse of a dominant position and, therefore, infringed Article 102 TFEU.

28

On 27 April 2015, the Commission granted Google access to the file.

29

Between June and September 2015, the Commission sent a non-confidential version of the statement of objections to 24 complainants and 10 interested parties. Comments were submitted by 20 complainants and 7 interested parties.

30

On 27 August 2015, Google submitted its response to the statement of objections.

31

Between October and November 2015, the Commission sent a non-confidential version of the response to the statement of objections to 23 complainants and 9 interested parties. Comments were submitted by 14 complainants and 7 interested parties.

32

On 14 July 2016, the Commission adopted a supplementary statement of objections.

33

On 27 July 2016, the Commission granted Google further access to the file.

34

Between September and October 2016, the Commission sent a non-confidential version of the supplementary statement of objections to 20 complainants and 6 interested parties. Comments on the supplementary statement of objections were submitted by 9 complainants and 3 interested parties.

35

On 3 November 2016, Google submitted its response to the supplementary statement of objections.

36

On 28 February 2017, the Commission sent Google a ‘letter of facts’, drawing its attention to evidence that was not expressly relied on in the statement of objections and the supplementary statement of objections, but which, on further analysis of the file, could be potentially relevant to support the preliminary conclusion reached in those documents.

37

On 1 March 2017, the Commission granted Google further access to the file.

38

On 18 April 2017, Google replied to the ‘letter of facts’.

39

On 27 June 2017, the Commission adopted Decision C(2017) 4444 final relating to proceedings under Article 102 TFEU and Article 54 of the EEA Agreement (Case AT.39740 – Google Search (Shopping)) (‘the contested decision’).

40

In the contested decision, after setting out the stages of the procedure leading to its adoption and rejecting Google’s claims concerning the conduct of that procedure, the Commission first of all defined the relevant markets, within the meaning of the competition rules.

41

The Commission recalled that, when investigating the possible dominant position of an undertaking on a market, it was required to take into consideration not only the characteristics of the products or services concerned, but also the structure of supply and demand, in order to determine the relevant market or markets. It stated that the distinctness of products or services in that context had to be assessed by reference to customer demand.

42

The Commission concluded that the relevant product markets were the market for online general search services and the market for online comparison shopping services.

43

In the first place, as regards the market for online general search services, the Commission stated that the provision of such a service was an economic activity because even though internet users use the service free of charge, they agree to allow the search engine operator to collect data concerning them, which it may subsequently monetise, particularly with advertisers wishing to display advertisements on the results pages. Generally, on ‘two-sided’ platforms, one side being free of charge for one user group (in this case, internet users) makes it possible, if the platform functions well, to strengthen demand for the other side, whose user group (in this case, advertisers who want to reach as many internet users as possible) is required to pay. To that extent, the various online general search services compete to attract both internet users and advertisers through the quality of their search engine.

44

Next, the Commission found that, from the standpoint of internet users’ demand, there was limited substitutability between general search services and other internet services.

45

In particular, according to the Commission, there is limited substitutability between specialised search services and general search services, since the former cover only their respective fields of specialisation. Moreover, specialised search services mostly provide commercial offerings only, whereas general search services provide all types of online service. The way in which the responses of the various search services are returned is also different, even if only in terms of the composition of their databases. Their financial models also differ, in that general search services are financed solely by payments for the display of advertisements on the results pages, while specialised search services are additionally financed by payments from undertakings whose websites are mentioned in the search results when internet users take follow-up action (payments linked to clicks or subsequent transactions). Specific examples, such as Google, confirm those distinctions. Thus, many undertakings offering specialised search services, such as Shopzilla (a comparison shopping service) or Kayak (a travel fare comparison service), do not offer a general search service. Google itself clearly distinguishes between the two types of search service and, as a matter of course, has specific search pages and results pages for its specialised search services. Industry analysts also draw a distinction between the two types of service. The Commission draws attention to further distinctions concerning the functionalities or use of both types of service, even though they may sometimes provide responses to the same query.

46

As regards supply side substitutability, the Commission also indicated that there was limited substitutability between general search services and other online services. In that regard, it cited the existence of barriers to entry in the case of general search services for operators of other online services to demonstrate that it would be difficult for them, in the short term and without incurring significant costs or risks, to compete with existing providers of general search services.

47

In essence, according to the Commission, a provider of online services wishing to offer a new general search service would have to make very substantial investments. A number of major internet companies reported the existence of serious barriers to entry in that respect. If a general search service is to function smoothly and be viable, it needs to receive a significant volume of search queries. Since the quality of the responses to internet users’ queries has undergone considerable change, a shift in market positions of the kind witnessed in the past, when Google overtook the former leading search engines Alta Vista and Lycos, is no longer likely today. The development of advertising on the general results pages also favours the market leader which attracts more advertisers given the number of users using its general search service. This makes it all the more difficult for new operators to emerge and it has on the contrary been observed that, since 2007, a number of operators have abandoned the business or confined themselves to a particular national market or language area. Only Microsoft has been able to pursue that business in any meaningful way with its search engine Bing. However, its market share does not exceed 10% in any EEA country.

48

Next, the Commission found that online general search services should not be distinguished according to whether internet users use them on computers or on other devices such as tablets or smartphones. It thus concluded that there was a product market for online general search services.

49

In the second place, as regards the market for online comparison shopping services, the Commission gave the following reasons for its existence. Comparison shopping services can be distinguished from other specialised search services on the internet. From the demand side perspective, each specialised search service deals with queries focusing on a specific subject matter and provides answers on that subject matter alone, so that there is no substitutability between the different specialised search services. From the supply side perspective, the criteria for selecting answers, the content of databases, the nature and sphere of activity of the operators of websites to which a specialised search service may direct users and the contractual relationships with those operators are so varied depending on the type of specialised search involved that it would be difficult for the provider of a specialised search service to offer, in the short term and without incurring significant additional costs, a different type of specialised search service and therefore to compete in that respect. Accordingly, supply side substitutability does not exist either between the different types of specialised search service.

50

For various reasons, there is also limited substitutability between services for the display of general advertisements on the general results pages (referred to as ‘online search advertising platforms’ in the contested decision) and comparison shopping services. The Commission put forward reasons relating essentially to the development and functioning of the two types of service, particularly the fact that internet users do not specifically look for advertising, whereas they deliberately turn to a comparison shopping service for results.

51

There is also limited substitutability between the services of online direct sellers and comparison shopping services. The Commission pointed out, in essence, that direct sellers focus on the products or services which they themselves sell and that the fact that internet users can purchase an item from them without having a comparison shopping service run a search did not mean that there was substitutability between the two types of service, which are very different.

52

There is limited substitutability between the services of online sales platforms, known as ‘merchant platforms’, and comparison shopping services. In response in particular to a number of arguments to the contrary put forward by Google, the Commission presented a detailed analysis of what it considered to be the differences between the two types of service, relating particularly to the fact that those two types of service, in its view, met different needs of internet users and online sellers, including in so far as, unlike online sales platforms, comparison shopping services did not sell products and thus did not provide services or assume obligations linked to the sale.

53

Concerning the geographic scope of the relevant markets, the Commission concluded that both the markets for general search services and the markets for specialised comparison shopping search services were national in scope. Even though websites can be accessed anywhere, factors related to national partitioning, particularly of a linguistic nature, and the existence of ‘national’ search engines led to that conclusion, which, moreover, Google does not dispute.

54

Next, the Commission stated that, since 2008, Google has held a dominant position on the market for general search services in every EEA country except the Czech Republic, where it has held such a position only since 2011. The Commission relied on a number of factors in that respect. It drew attention to Google’s very high and stable market shares by volume, as observed in various studies, which have almost always exceeded 80% since 2008, except in the Czech Republic, where Google nevertheless became the undisputed market leader in January 2011 with a market share exceeding 70%. The Commission pointed to the low market shares of Google’s competitors, such as Bing and Yahoo. It restated the considerations concerning barriers to market entry set out in its earlier analysis of the market definition, and also stated that few internet users used more than one general search engine, that Google had a strong reputation and that internet users, being independent of each another, did not exert any countervailing buyer power. The Commission rejected Google’s arguments that the fact that its service was offered to internet users free of charge changed the situation, and stated that Google’s dominant position existed in relation to searches carried out using both desktop computers and mobile devices.

55

The Commission then found that Google had, at different times dating back as far as January 2008, abused the dominant position it held in 13 national markets for general search services within the EEA by decreasing traffic from its general results pages to competing comparison shopping services and increasing traffic to its own comparison shopping service, which was capable of having, or was likely to have, anticompetitive effects on the 13 corresponding national markets for specialised comparison shopping search services and on those national markets for general search services. The countries concerned are Belgium, the Czech Republic, Denmark, Germany, Spain, France, Italy, the Netherlands, Austria, Poland, Sweden, the United Kingdom and Norway.

56

The Commission described the abuse that Google was alleged to have committed as follows. As regards the principles at issue, the Commission stated that the prohibitions in Article 102 TFEU and Article 54 of the EEA Agreement could cover not only the conduct of an undertaking that was tending to strengthen its position on the market on which it was already dominant, but also the conduct of an undertaking in a dominant position on a given market that was tending to extend its position to a neighbouring market by distorting competition. The Commission recalled that an abuse of a dominant position was prohibited regardless of the means and procedure by which it was achieved and irrespective of any fault. Nevertheless it was open to the undertaking concerned to provide a justification by demonstrating that its conduct was objectively necessary or that the exclusionary effect produced could be counterbalanced by advantages in terms of efficiency gains that also benefited consumers.

57

The Commission stated, in Section 7.2 of the contested decision, that the abuse identified in the present case consisted in the more favourable positioning and display, in Google’s general results pages, of its own comparison shopping service compared to competing comparison shopping services.

58

In order to demonstrate why those practices were abusive and fell outside the scope of competition on the merits, in the first place, the Commission described, in Section 7.2.1 of the contested decision, how Google positioned and displayed its own comparison shopping service more favourably than competing comparison shopping services. The Commission examined, first of all, how competing comparison shopping services were positioned and displayed among Google’s generic results, before going on to examine how Google’s comparison shopping service was positioned and displayed within its general results pages.

59

As regards, first, the positioning of competing comparison shopping services, the Commission observed that they appeared in the generic results, in the form of links to their results pages capable of answering a user’s query. At the same time they were prone to being demoted within the ranking of generic results due to the application of ‘adjustment’ algorithms, in particular the ‘Panda’ algorithm, on account of, inter alia, the characteristics of the comparison shopping services and especially their lack of original content. The Commission stated, among other things, that, since their rollout, the algorithms in question had been applied to the great majority of the 361 comparison shopping services identified by Google in its response to the statement of objections (‘the 361 competing comparison shopping services identified by Google’) and that, in the United Kingdom, Germany, France, Italy and Spain, in the period between 2 August 2010 and 2 December 2016, the visibility of competing comparison shopping services on Google’s general results pages, which was at its highest at the end of 2010 and the beginning of 2011, had suffered a sudden drop after the launch of the Panda algorithm and never recovered.

60

As regards, secondly, the display of competing comparison shopping services, the Commission noted that these could only be displayed as generic results on Google’s general results pages, that is to say, in the form of simple blue links, and could not, therefore, be displayed in rich format with images and additional information on the products, prices and seller, although such information increased the click-through rate. The Commission mentioned certain evidence in support of that assertion, including studies and experiments.

61

Next, the Commission examined how Google’s comparison shopping service was positioned and displayed on the general results pages. As regards its positioning, the Commission identified two differences with respect to the positioning of competing comparison shopping services: first, Google’s comparison shopping service was not subject to the same ranking mechanisms, particularly the adjustment algorithms such as Panda, and, secondly, when Google’s comparison shopping service was displayed in a ‘box’, it appeared in a highly visible place. Concerning the application of adjustment mechanisms, the Commission noted that those algorithms did not apply to Google’s comparison shopping service despite the fact that it had numerous characteristics in common with competing comparison shopping services that would have made it prone to the same demotions in the generic results. So far as the visibility of Google’s comparison shopping service in the general results pages was concerned, the Commission stated more specifically that, since the launch of the Product Universal, Google had in most cases positioned the results of its own comparison shopping service either above all the generic results or at the level of the first generic results, the objective being, according to an internal Google email, to ‘dramatically increase traffic’. After describing the evolution of the Product Universal between 2007 and 2012, the Commission examined the positioning of the Shopping Unit and found that it was always positioned above Google’s first generic results. In response to Google’s argument that the trigger rate of the Shopping Unit was low, the Commission pointed out that, in most cases, the trigger rate of the Shopping Unit exceeded the trigger rate of the 361 competing comparison shopping services identified by Google, both in the first four generic results and as the first generic result. In support of that assertion, the Commission provided figures for the 13 geographic markets at issue.

62

As regards the display of Google’s comparison shopping service, the Commission found that the main difference in display compared to competing comparison shopping services was that Google’s comparison shopping service was displayed with richer graphical features, including images and dynamic information. According to the Commission, those richer graphical features led to higher click-through rates for Google and therefore to an increase in its revenue. The Commission listed several reasons in support of that assertion, based on Google’s own explanations and on another undertaking’s submissions in the administrative procedure.

63

Next, the Commission replied to the arguments put forward by Google to challenge the claim that it engaged in favouring. In particular, the Commission set out various reasons why the display and use of Product Universals and Shopping Units favoured Google’s comparison shopping service. It also considered the argument that Google applied the same relevance standards (i) to the Product Universal and to generic results, and (ii) to the Shopping Unit and to other product ads, to be irrelevant.

64

In order to demonstrate the abusive nature of the practices at issue, in the second place, the Commission examined in Section 7.2.2 of the contested decision the value of traffic volume for comparison shopping services. The Commission noted that the volume of traffic was important in many respects for the ability of a comparison shopping service to compete. After quoting the owner of several comparison shopping services on that point, according to whom traffic is the most important asset of a specialised search engine because, for a variety of reasons, the greater the traffic, the greater the relevance of search services, the Commission confirmed in particular, on the basis of numerous statements, that the relevance of a specialised search service was related to the breadth and freshness of the information provided. A significant volume of traffic enabled comparison shopping services to convince sellers to provide them with more data on their products, thus increasing the online comparison shopping services they offer and, in turn, their revenue. The Commission also noted, quoting numerous statements in that regard, that traffic led to machine learning effects, thereby improving the relevance of the search results and thus the usefulness of the comparison shopping service offered to internet users. Lastly, the Commission explained that traffic allowed comparison shopping services to carry out experiments aimed at improving their search services and suggesting additional searches to internet users who consult them.

65

In order to demonstrate the abusive nature of the practices at issue, in the third place, the Commission explained in Section 7.2.3 of the contested decision that those practices decreased traffic from Google’s general results pages to competing comparison shopping services and increased traffic from those pages to Google’s comparison shopping service. The Commission gave three reasons to support that finding. First of all, on the basis of an analysis of internet users’ behaviour, the Commission concluded that generic results generated significant traffic to a website when they were ranked within the first three to five results on the first general results page (‘above the fold’), internet users paying little or no attention to subsequent results, which often did not appear directly on the screen. The Commission added that the first 10 results received approximately 95% of internet users’ clicks. On the basis of studies conducted by Microsoft, the Commission specified that the position of a given link in the generic results had a major impact on the click-through rate of that link, irrespective of the relevance of the web page to which it led, and that a change in the ranking of a search result on Google’s general results pages had a major impact on traffic flowing from the general search. Next, the Commission stated that the practices at issue had led to a decrease in traffic from Google’s general results pages to almost all competing comparison shopping services over a significant period of time in each of the 13 EEA countries where those practices had been implemented. Lastly, the Commission found that the practices at issue had led to an increase in Google’s traffic to its own comparison shopping service. The Commission relied on various items of evidence to support those findings. It contested the arguments which Google had put forward to challenge the traffic trends identified or the causal link between its conduct and those trends.

66

In order to demonstrate the abusive nature of the practices at issue, in the fourth place, the Commission claimed, in Section 7.2.4 of the contested decision, that the traffic diverted by those practices accounted for a large proportion of traffic to competing comparison shopping services and that it could not be effectively replaced by other sources of traffic currently available to competing comparison shopping services, namely AdWords text ads, mobile applications, direct traffic, referrals from other partner websites, social networks and other general search engines.

67

In order to demonstrate the abusive nature of the practices at issue, in the fifth place, the Commission explained in Section 7.3 of the contested decision that those practices had potential anticompetitive effects on the 13 national markets for specialised comparison shopping search services and on the 13 national markets for general search services mentioned in paragraph 55 above. With regard to the national markets for specialised comparison shopping search services, the Commission sought to demonstrate that the practices at issue could cause competing comparison shopping services to cease trading, have a negative impact on innovation and therefore reduce the ability of consumers to access the most relevant services. The competitive structure of those markets would thus be affected. If merchant platforms were to be included in those markets, the Commission considered that the same effects would be felt by Google’s closest competitors, namely competing comparison shopping services. As regards the national markets for general search services, according to the Commission, the anticompetitive effects of the practices at issue arise from the fact that the additional resources generated by Google’s comparison shopping service from its general results pages enable it to strengthen its general search service.

68

In summary, in the contested decision, the Commission sought to demonstrate that Google was positioning and promoting its comparison shopping service on its general results pages more favourably than competing comparison shopping services (Section 7.2.1 of the contested decision); that significant traffic, in other words, a high number of visits, was essential for comparison shopping services (Section 7.2.2 of the contested decision); that Google’s conduct increased traffic to its comparison shopping service and decreased traffic to competing comparison shopping services (Section 7.2.3 of the contested decision); that traffic from Google’s general results pages accounted for a large proportion of the traffic of those competing comparison services and could not be effectively replaced by other sources of traffic (Section 7.2.4 of the contested decision); that the conduct at issue could result in Google’s dominant position being extended to markets other than the market on which that position was already held, namely the markets for specialised comparison shopping search services (Section 7.3.1 of the contested decision); that even if comparison shopping services were included in wider markets also encompassing the services of online sales platforms, the same anticompetitive effects would be felt in the segment of those markets covering comparison shopping services (Section 7.3.2 of the contested decision); and that that conduct also protected Google’s dominant position on the markets for general search services (Section 7.3.3 of the contested decision). In particular, the Commission drew attention to the harm that could be caused to consumers as a result of the situation. It contested the arguments put forward by Google in challenging that analysis, to the effect that the legal criteria used were wrong (Section 7.4 of the contested decision). The Commission also rejected the reasons put forward by Google to demonstrate that its conduct was not abusive (Section 7.5 of the contested decision), whereby Google claimed that it was objectively necessary or that any resulting restrictions of competition were offset by efficiency gains benefiting consumers.

69

As is apparent in particular from recitals 344 and 512 of the contested decision, the conduct specifically identified by the Commission as the source of Google’s abuse is, in essence, the fact that Google displayed its comparison shopping service on its general results pages in a prominent and eye-catching manner in dedicated ‘boxes’, without that comparison service being subject to the adjustment algorithms used for general searches, whereas, at the same time, competing comparison shopping services could appear on those pages only as general search results (blue links) that tended to be given a low ranking as a result of the application of those adjustment algorithms. The Commission pointed out, in recitals 440 and 537 of the contested decision, that it did not object, per se, to the various selection criteria chosen by Google, described as relevance criteria, but to the fact that the same positioning and display criteria were not applied both to Google’s own and to competing comparison shopping services. Similarly, in recital 538 of the contested decision, the Commission stated that it did not object, as such, to the promotion of specialised comparison shopping results that Google considered to be relevant, but to the fact that the same promotion effort was not made in respect of both Google’s own comparison shopping service and competing comparison shopping services.

70

After setting out the above evidence, the Commission declared, in Article 1 of the contested decision, that Google Inc. and Alphabet, since its takeover of Google Inc., had infringed Article 102 TFEU and Article 54 of the EEA Agreement in the 13 countries mentioned in paragraph 55 above, which were either Member States of the European Union or other States party to the EEA Agreement, from various dates corresponding to the introduction of specialised product results or product ads on Google’s general results page.

71

The Commission considered that the situation was such that Google should be ordered to bring an end to the conduct at issue within 90 days and to refrain from similar conduct having the same object or effect. It made clear that although Google could comply with that order in different ways, certain principles had to be respected, regardless of whether or not Google decided to retain Shopping Units or other groups of comparison shopping search results on its general results pages. Those principles included, in essence, the principle of non-discrimination between Google’s comparison shopping service and competing comparison shopping services. The order requiring Google to bring an end to the conduct at issue appears in Article 3 of the operative part of the contested decision.

72

Lastly, the Commission considered that a pecuniary penalty should be imposed on Google. It recalled that, under Article 23(2)(a) of Regulation No 1/2003 and Article 5 of Council Regulation (EC) No 2894/94 of 28 November 1994 concerning arrangements for implementing the Agreement on the European Economic Area (OJ 1994 L 305, p. 6), it could impose such a penalty on undertakings which had, either intentionally or negligently, infringed Article 102 TFEU and Article 54 of the EEA Agreement. It also recalled the general parameters for determining pecuniary penalties set out in Article 23(3) of Regulation No 1/2003, namely the gravity and duration of the infringement, and the way in which it had indicated those parameters would be applied, in its Guidelines on the method of setting fines imposed pursuant to Article 23(2)(a) of Regulation No 1/2003 (OJ 2006 C 210, p. 2; ‘the Guidelines’).

73

The Commission found that Google could not have been unaware of its dominant position on the national markets for general search services or of the abusive nature of its conduct, even though some aspects of the situation had not been examined in previous cases. Google had therefore acted intentionally or negligently. The Commission considered that the fact that discussions had been held at one stage in the procedure to address the competition issue identified by means of commitments to be given by Google did not preclude the imposition of a fine.

74

The Commission then stated that, in view of the control exercised by Alphabet over Google Inc. since 2 October 2015, Alphabet was to be jointly and severally liable for payment of the fine imposed in so far as it related to the period from that date.

75

Next, the Commission determined that the basic amount as a starting point for calculating the pecuniary penalty, defined in points 12 to 19 of the Guidelines as the ‘value of sales’, would be the revenue generated in 2016, in the 13 countries in which it had identified the conduct at issue, by product ads appearing in Shopping Units or on the specialised Google Shopping page and by text ads also appearing on that page.

76

The Commission concluded that, in view of the economic importance of the 13 national markets for comparison shopping services and the fact that Google not only held a dominant position in the countries concerned on the market for general search services, but was also far ahead of its competitors in terms of market shares, the gravity coefficient to be used to determine the pecuniary penalty, as provided for in points 20 to 23 of the Guidelines, had to be 10% of the basic amount described in paragraph 75 above. As provided for in point 24 of the Guidelines, the Commission then, for each of the 13 countries in which an infringement had been found, multiplied that amount by the number of years of infringement that had elapsed since the launch of the Product Universal or, failing that, of the Shopping Unit. On that basis, the Commission found periods lasting from 1305 to 3435 days, depending on the country.

77

In order to ensure, in essence, that the penalty had a deterrent effect, including on undertakings of a similar size and with a similar financial capacity to Google – noting that its overall turnover was EUR 81597000000 in 2016 – the Commission added an additional amount, as provided for in point 25 of the Guidelines, corresponding to 10% of the basic amount referred to in paragraph 75 above, and multiplied the resulting figure by 1.3. It did not find that there were any aggravating or mitigating circumstances that would have justified an increase or decrease in the amount of the fine.

78

Thus, by Article 2 of the contested decision, the Commission imposed on Google Inc. a pecuniary penalty of EUR 2424495000, of which EUR 523518000 jointly and severally with Alphabet.

79

By application lodged at the Court Registry on 11 September 2017, Google brought the present action.

80

By document lodged at the Court Registry on 28 November 2017, the Bureau européen des unions de consommateurs (BEUC) applied for leave to intervene in support of the form of order sought by the Commission.

81

By document lodged at the Court Registry on 4 December 2017, Connexity Inc., Connexity UK Ltd, Connexity Europe GmbH and Pricegrabber.com applied for leave to intervene in support of the form of order sought by the Commission.

82

By document lodged at the Court Registry on 7 December 2017, Infederation Ltd (‘Foundem’) applied for leave to intervene in support of the form of order sought by the Commission.

83

By documents lodged at the Court Registry on 11 December 2017, the EFTA Surveillance Authority and Initiative for a Competitive Online Marketplace applied for leave to intervene in support of the form of order sought by the Commission.

84

By document lodged at the Court Registry on 19 December 2017, Prestige Gifting Ltd applied for leave to intervene in support of the form of order sought by Google.

85

By document lodged at the Court Registry on 19 December 2017, Kelkoo applied for leave to intervene in support of the form of order sought by the Commission.

86

By document lodged at the Court Registry on 20 December 2017, Computer & Communication Industry Association (‘CCIA’) applied for leave to intervene in support of the form of order sought by Google.

87

By documents lodged at the Court Registry on 20 December 2017, Consumer Watchdog, Yelp Inc., Verband Deutscher Zeitschriftenverleger eV (‘VDZ’), Visual Meta GmbH, BDZV – Bundesverband Digitalpublisher und Zeitungsverleger eV, formerly Bundesverband Deutscher Zeitungsverleger eV (‘BDZV’), the Federal Republic of Germany, Open Internet Project (OIP) and Twenga applied for leave to intervene in support of the form of order sought by the Commission.

88

By document lodged at the Court Registry on 21 December 2017, FairSearch applied for leave to intervene in support of the form of order sought by the Commission.

89

The Commission lodged the defence on 31 January 2018.

90

By document lodged at the Court Registry on 20 March 2018, StyleLounge GmbH applied for leave to intervene in support of the form of order sought by the Commission.

91

By letter of 23 March 2018, Google and the Commission requested, pursuant to Article 144 of the Rules of Procedure of the General Court, that certain information in the case file not be communicated to the interveners owing to its confidential nature. The requests submitted by Google and the Commission were identical in content in that respect with regard to all of the applicants for leave to intervene, including the EFTA Surveillance Authority.

92

Google lodged the reply on 7 May 2018.

93

By order of 16 May 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:292), the President of the Ninth Chamber of the General Court dismissed the application by StyleLounge to intervene in the proceedings in support of the Commission as it was out of time.

94

The Commission lodged the rejoinder on 20 July 2018.

95

Following a measure of organisation of procedure adopted by the Court with a view to reducing the scope of the applications for confidential treatment of information in the case file, Google and the Commission submitted, in relation to all of the applicants for leave to intervene, revised requests for confidential treatment concerning the application and the defence on 28 September 2018 and, subsequently, requests for confidential treatment concerning the reply and the rejoinder on 12 October 2018. Those requests were also identical in content with regard to all of the applicants for leave to intervene.

96

By orders of 7 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:978), of 7 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:982), of 7 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:996), of 7 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1001), and of 7 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1002), the President of the Ninth Chamber of the General Court dismissed the applications to intervene lodged by, respectively, Prestige Gifting, FairSearch, Consumer Watchdog, Yelp, Connexity, Connexity UK, Connexity Europe and Pricegrabber.com and Initiative for a Competitive Online Marketplace for failure to establish an interest in the result of the case.

97

By orders of 17 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1007), of 17 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1008), of 17 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1009), of 17 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1010), of 17 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1011), of 17 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1028), and of 17 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1029), the President of the Ninth Chamber of the General Court granted leave to intervene, respectively, to BEUC, Foundem, CCIA, VDZ, BDZV, Visual Meta, Twenga, the EFTA Surveillance Authority, Kelkoo and the Federal Republic of Germany. In those orders, the costs related to the interventions were reserved.

98

By order of 17 December 2018, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2018:1005), the President of the Ninth Chamber of the General Court dismissed OIP’s application to intervene for failure to establish an interest in the result of the case.

99

In the orders granting leave to intervene, the decision as to the merits of the requests for confidential treatment was reserved and a non-confidential version of the procedural documents was communicated to BEUC, Foundem, CCIA, VDZ, BDZV, Visual Meta, Twenga, the EFTA Surveillance Authority, Kelkoo and the Federal Republic of Germany pending the submission of any observations on their part on the requests for confidential treatment.

100

On 15 January 2019, Foundem contested in part Google’s requests for confidential treatment.

101

On 15 January 2019, and subsequently on 25 January 2019, the EFTA Surveillance Authority indicated that, so far as it was concerned, the requests for confidential treatment of the Commission and of Google were, in whole or in part, devoid of purpose or unfounded. It nevertheless made clear that it was not asking to be provided with the confidential versions of the documents in the file.

102

By order of 11 April 2019, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2019:250), the President of the Ninth Chamber of the General Court granted some of the revised requests for confidential treatment concerning information in the application and the defence and some of the requests for confidential treatment concerning information in the reply and the rejoinder. It refused the requests for confidential treatment as to the remainder. Consequently, a time limit was set for Google and the Commission to submit new non-confidential versions of certain documents in the file, and a time limit was set for Foundem to supplement its statement in intervention in the light of the information that was no longer treated as confidential. In response to the observations of the EFTA Surveillance Authority, which had invoked its special position in administrative procedures leading to Commission decisions, such as the contested decision, finding an infringement of the competition rules laid down in the EEA Agreement, the President of the Ninth Chamber of the General Court stated that, in the context of judicial proceedings before the Court, that authority was subject to the same requirements as the other interveners and that, in the circumstances of the case, the observations of the EFTA Surveillance Authority could not be acted upon.

103

BEUC, Foundem, CCIA, VDZ, BDZV, Visual Meta, Twenga, the EFTA Surveillance Authority, Kelkoo and the Federal Republic of Germany each lodged their statement in intervention on 15 March 2019 and Foundem lodged a supplementary statement in intervention on 11 June 2019. The Commission submitted observations on CCIA’s statement in intervention on 20 May 2019 and Google submitted observations on the statements in intervention of BEUC, CCIA, VDZ, BDZV, Visual Meta, Twenga, the EFTA Surveillance Authority, Kelkoo and the Federal Republic of Germany, but not of Foundem, on 21 June 2019 and, specifically as regards the statement in intervention of Foundem, on 1 July 2019.

104

Acting on a proposal from the Ninth Chamber, the Court decided on 10 July 2019, pursuant to Article 28 of the Rules of Procedure, to refer the case to a Chamber sitting in extended composition.

105

By letters of 9 and 23 August 2019, the Commission and Google respectively requested that, owing to their confidential nature, certain elements of Google’s observations on a number of statements in intervention not be communicated to BEUC, Foundem, CCIA, VDZ, BDZV, Visual Meta, Twenga, the EFTA Surveillance Authority, Kelkoo and the Federal Republic of Germany.

106

On 5 and 10 September 2019, BEUC and Kelkoo respectively challenged certain requests for confidentiality made by Google concerning its observations on their statements in intervention.

107

By order of 8 October 2019, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2019:770), the President of the Ninth Chamber (Extended Composition) of the General Court held that there was no need to adjudicate on the uncontested requests for confidentiality referred to in paragraph 105 above and, as regards those that were contested, granted certain requests as regards BEUC, CCIA, VDZ, BDZV, Visual Meta, Twenga, the EFTA Surveillance Authority, Kelkoo and the Federal Republic of Germany, granted certain other requests save with regard to Kelkoo and rejected others.

108

Acting on a report of the Judge-Rapporteur, the General Court (Ninth Chamber, Extended Composition) decided to open the oral part of the procedure and, pursuant to Article 89(2) and (3) of the Rules of Procedure, invited the main parties to reply to a number of questions, either in writing or at the hearing.

109

On 21 and 22 January 2020, the Commission and Google respectively replied to the questions put by the Court that required a written response. Google requested that, owing to their confidential nature, certain elements of its response not be communicated to BEUC, Foundem, CCIA, VDZ, BDZV, Visual Meta, Twenga, the EFTA Surveillance Authority, Kelkoo and the Federal Republic of Germany.

110

On 5 February 2020, BDZV challenged certain requests for confidentiality made by Google concerning its written reply to the questions put by the Court, but also concerning annexes to the defence and to the reply.

111

By order of 10 February 2020, Google and Alphabet v Commission (T‑612/17, not published, EU:T:2020:69), the President of the Ninth Chamber (Extended Composition) of the General Court dismissed as inadmissible the applications by BDZV for confidentiality to be waived in respect of annexes to the defence and to the reply, held that there was no need to adjudicate on the uncontested requests for confidentiality referred to in paragraph 109 above, and, as regards those that were contested, granted them.

112

The hearing was held from 12 to 14 February 2020, after the main parties had agreed to waive confidentiality with regard to BEUC, Foundem, CCIA, VDZ, BDZV, Visual Meta, Twenga, the EFTA Surveillance Authority, Kelkoo, the Federal Republic of Germany and the public in respect of certain elements of the file, following a preparatory meeting of the President of the Chamber and the Judge-Rapporteur with the main parties that took place on 15 January 2020 on the basis of Article 89 of the Rules of Procedure.

113

Google claims that the Court should:

114

The Commission contends that the Court should:

115

CCIA claims that the Court should annul the contested decision and order the Commission to pay the costs relating to CCIA’s intervention.

116

The Federal Republic of Germany contends that the Court should dismiss the action.

117

BEUC contends that the Court should dismiss the action and order Google to pay the costs relating to BEUC’s intervention.

118

Foundem, Kelkoo, VDZ, Visual Meta, BDZV and Twenga contend that the Court should dismiss the action and order Google to pay the costs.

119

It must be emphasised at the outset that Google does not dispute the fact that it holds a dominant position on the 13 national markets for general search services corresponding to the countries in which the Commission found that Google had abused that position. That fact is a premiss on which all the analyses that follow are based.

120

Google raises six pleas for annulment of the contested decision, which it presents as follows:

‘The First and Second pleas show that the Decision errs in finding that Google favours a Google comparison shopping service by showing Product Universals and Shopping Units. The Third plea explains that the Decision errs in finding that the positioning and display of Product Universals and Shopping Units diverted Google search traffic. The Fourth plea demonstrates that the Decision’s speculation about anticompetitive effects is unfounded. The Fifth plea shows that the Decision errs in law by treating quality improvements that constitute competition on the merits as abusive. The Sixth plea sets out why the Decision errs in imposing a fine.’

121

The Court observes that Google’s arguments contain numerous factual and technical elements and criticisms of a legal nature that are restated in support of various pleas. The Court will address Google’s pleas and arguments in the following order.

122

The Court will first of all, in Section B of this part of the present judgment, concerning the principal claim, examine in point 1 Google’s arguments that the practices with which the Commission takes issue are in fact quality improvements in its online search service (fifth plea), from which it follows (i) that Google could not have committed an abuse, the Commission having failed to identify any elements of those improvements that represent a departure from competition on the merits, and (ii) that, having been unable to isolate those elements, the Commission in fact imposed on Google a duty to supply, without satisfying the strict conditions laid down by the judgment of 26 November 1998, Bronner, (C‑7/97, EU:C:1998:569). More specifically, according to Google, the Commission required it to give its competitors access to its services as though these were an ‘essential facility’ that was indispensable to them, without demonstrating that the requisite conditions laid down by the case-law of the Court of Justice were satisfied. Also in the context of Google’s arguments concerning competition on the merits, the Court will examine the argument that, in essence, Google did not have an anticompetitive objective in implementing the specialised results at issue in the present case, which constitute quality improvements in its search service. That argument, raised in the first part of the first plea, will also be examined in Section B.1.

123

Thus, the Court will examine, in Section B.1, the legality of the legal classification of favouring adopted by the Commission on the basis of Article 102 TFEU and the question whether such a concept of abuse, by which an undertaking in a dominant position is, in essence, alleged to be favouring its own service at the expense of its competitors’ services, could lawfully be accepted by the Commission.

124

Next, the Court will examine, in Section B.2 of this part of the present judgment, whether there is in fact a difference in treatment underpinning that classification, namely whether or not Google discriminated in favour of its own specialised search service (i) in the period during which the Product Universal was in place (first plea) and (ii) in the period during which the Shopping Unit was in place (second plea).

125

In Section B.3 of this part of the present judgment, the Court will then examine Google’s third and fourth pleas, according to which the conduct at issue did not have anticompetitive effects.

126

Lastly, in Section B.4 of this part of the present judgment, the Court will examine the third parts of the first and second pleas submitted by Google, according to which the conduct at issue was objectively justified and, consequently, was not contrary to Article 102 TFEU.

127

After having drawn a conclusion on the merits in Section B.5 of this part of the present judgment, the Court will, in Section C, examine the sixth plea submitted by Google, according to which the pecuniary penalty is in any event unjustified and, at the very least, too high.

128

As a preliminary point, the Court recalls the extent of review by the Courts of the European Union of decisions adopted by the Commission under Article 102 TFEU.

129

Judicial review by the General Court consists in a review of the legality of the acts of the institutions for which provision is made in Article 263 TFEU, which may be supplemented, pursuant to Article 261 TFEU, by unlimited jurisdiction with regard to the penalties imposed (see, to that effect, judgment of 21 January 2016, Galp Energía España and Others v Commission, C‑603/13 P, EU:C:2016:38, paragraph 71).

130

As the Court of Justice has stated, the scope of judicial review provided for in Article 263 TFEU extends to all the elements of Commission decisions relating to proceedings applying Articles 101 and 102 TFEU which are subject to in-depth review by the General Court, in law and in fact, in the light of the pleas raised by the applicants and taking into account all the elements submitted by the latter, whether those elements pre-date or post-date the contested decision, whether they were submitted previously in the context of the administrative procedure or, for the first time, in the context of the proceedings before the General Court, in so far as those elements are relevant to the review of the legality of the Commission decision (judgment of 21 January 2016, Galp Energía España and Others v Commission, C‑603/13 P, EU:C:2016:38, paragraph 72; see also, to that effect, judgment of 26 September 2018, Infineon Technologies v Commission, C‑99/17 P, EU:C:2018:773, paragraph 48).

131

The Court of Justice has held that while the Commission has a margin of discretion with regard to economic matters, that does not mean that the Courts of the European Union must refrain from reviewing the Commission’s interpretation of information of an economic nature. Those Courts must, among other things, not only establish whether the evidence put forward is factually accurate, reliable and consistent, but must also determine whether that evidence contains all the relevant data that must be taken into consideration in appraising a complex situation and whether it is capable of substantiating the conclusions drawn from it (judgments of 15 February 2005, Commission v Tetra Laval, C‑12/03 P, EU:C:2005:87, paragraph 39; of 8 December 2011, Chalkor v Commission, C‑386/10 P, EU:C:2011:815, paragraph 54; and of 10 July 2014, Telefónica and Telefónica de España v Commission, C‑295/12 P, EU:C:2014:2062, paragraph 54). Where, in order to classify a practice in the light of the provisions of Article 102 TFEU, the Commission attaches real importance to an economic analysis, the Courts of the European Union are required to examine all of the arguments put forward by the undertaking penalised concerning that analysis (see, to that effect, judgment of 6 September 2017, Intel v Commission, C‑413/14 P, EU:C:2017:632, paragraphs 141 to 144).

132

In addition, it is apparent from the case-law of the Court of Justice that, in the field of competition law, where there is a dispute as to the existence of an infringement, it is for the Commission to prove the infringements found by it and to adduce evidence capable of demonstrating to the requisite legal standard the existence of the circumstances constituting an infringement. Where the Court still has a doubt, the benefit of that doubt must be given to the undertakings accused of the infringement (judgments of 22 November 2012, E.ON Energie v Commission, C‑89/11 P, EU:C:2012:738, paragraphs 71 and 72, and of 16 February 2017, Hansen & Rosenthal and H&R Wax Company Vertrieb v Commission, C‑90/15 P, not published, EU:C:2017:123, paragraphs 17 and 18).

133

While it is for the authority alleging an infringement of the competition rules to prove it, it is for the undertaking raising a defence against the finding of an infringement of those rules to demonstrate that that defence must be upheld, so that the authority will then have to rely on other evidence in the contested decision. Furthermore, even if the burden of proof rests, according to those principles, either on the Commission or on the undertaking concerned, the evidence on which a party relies may be of such a kind as to require the other party to provide an explanation or justification, failing which it is permissible to conclude that the rules on the burden of proof have been satisfied (see, to that effect, judgment of 17 June 2010, Lafarge v Commission, C‑413/08 P, EU:C:2010:346, paragraphs 29 and 30 and the case-law cited).

134

Thus, when the Commission relies on evidence which is, in principle, sufficient to demonstrate the existence of the infringement, it is not sufficient that the undertaking concerned raises the possibility that a circumstance arose which might affect the probative value of that evidence so that the Commission bears the burden of proving that that circumstance was not capable of affecting the probative value of that evidence. On the contrary, except in cases where such proof could not be provided by the undertaking concerned on account of the conduct of the Commission itself, it is for the undertaking concerned to prove to the requisite legal standard, on the one hand, the existence of the circumstance relied on by it and, on the other, that that circumstance calls into question the probative value of the evidence relied on by the Commission (judgment of 22 November 2012, E.ON Energie v Commission, C‑89/11 P, EU:C:2012:738, paragraph 76).

135

Lastly, it should be noted that the Court of Justice and the General Court cannot under any circumstances, in the context of the review of legality referred to in Article 263 TFEU, substitute their own reasoning concerning the assessment of the facts for that of the author of the contested act (judgments of 27 January 2000, DIR International Film and Others v Commission, C‑164/98 P, EU:C:2000:48, paragraph 38; of 24 January 2013, Frucona Košice v Commission, C‑73/11 P, EU:C:2013:32, paragraph 89; and of 21 January 2016, Galp Energía España and Others v Commission, C‑603/13 P, EU:C:2016:38, paragraph 73). In so far as the review of the legality of the contested decision relates to the reasons stated in that decision, the Court cannot, either on its own initiative or at the request of the administration, add reasons to those given in that decision.

136

As indicated in paragraphs 122 and 123 above, first of all, Google asserts in the first part of the fifth plea in law that the contested decision fails to identify anything in Google’s conduct, which consisted in making quality improvements in its online search service, that would represent a departure from competition on the merits.

137

Next, Google claims, in the second part of the fifth plea, that the conduct at issue in the contested decision constitutes, in reality, a refusal to supply, since the Commission complains that Google failed to make its ‘technologies and designs’ and, in particular, the boxes at the top of its general results pages, accessible to results from competing comparison shopping services. In order to establish that such conduct was contrary to Article 102 TFEU, the Commission should have established that the conditions laid down in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), were satisfied, which it failed to do. By describing conduct as favouring, the Commission thus, in reality, sought to circumvent the conditions applicable to a refusal to supply, and its decision is accordingly vitiated by an error of law.

138

Lastly, Google submits in the first part of the first plea that the contested decision distorts the reasons for creating specialised product results. It claims that it did not introduce grouped product results in order to drive traffic to its own comparison shopping service, as alleged by the Commission, but to improve the quality of its results and their display for users.

139

In the first part of its fifth plea, Google claims that the practices at issue are quality improvements that constitute competition on the merits and cannot be treated as abusive.

140

On the first aspect, Google refers in particular to the judgments of 13 February 1979, Hoffmann-La Roche v Commission (85/76, EU:C:1979:36, paragraph 91), of 3 July 1991, AKZO v Commission (C‑62/86, EU:C:1991:286, paragraph 70), and of 14 October 2010, Deutsche Telekom v Commission (C‑280/08 P, EU:C:2010:603, paragraph 177), to explain that, so far as dominant undertakings are concerned, the Court of Justice distinguishes between anticompetitive abusive practices and pro-competitive conduct within the scope of ‘normal’ competition or competition ‘on the merits’.

141

Therefore, the critical point in all the cases giving rise to the judgments cited in paragraph 140 above is that the undertakings are entitled to use all ‘normal’ methods to compete to win business. That includes Google’s right to ‘compete better’ by improving the quality of its technologies and its specialised search services for the natural product results and product ads available via its general search page. CCIA submits in that regard that developing and improving the ‘design’ of a website is part of the competitive process and that these changes are what both consumers and advertisers have come to expect. The quality of a website is a key parameter of competition in online markets. CCIA adds that, in today’s economy, vertical integration is ubiquitous and generally a positive step from an economic perspective.

142

According to Google, the theory advanced in the contested decision identifies nothing that distinguishes its practices from competition on the merits. The claim that Google engaged in favouring and the presumption of potential effects do not change the fact that grouped product results and product ads have improved the quality of its general search service. By showing these ‘designs’ on its general results pages and developing the innovative technologies that supported them, Google competed on the merits in the market for general search services.

143

In Google’s submission, the Commission tried to sidestep the issues by arguing, in recital 334 of the contested decision, that the ‘conduct of an undertaking with a dominant position in a given market’ can be abusive if it ‘tends to extend that position to a neighbouring but separate market’. It claimed, in recital 652 of the contested decision, that applying that rule to a product or service improvement is in line with existing case-law. Thus, according to Google, the Commission merely considered that Google’s conduct was intended, through ‘leveraging’, to extend its dominant position to markets adjacent to those in which it held that position, but failed to take account of the fact that the conduct consisted in improving Google’s services and did not deviate from ‘normal’ competition or competition ‘on the merits’.

144

It is apparent from the case-law that not every exclusionary effect is necessarily detrimental to competition, since competition on the merits may lead to the disappearance or marginalisation of competitors that are less efficient. Google refers in that respect to the judgments of 27 March 2012, Post Danmark (C‑209/10, EU:C:2012:172, paragraph 22), and of 6 September 2017, Intel v Commission (C‑413/14 P, EU:C:2017:632, paragraph 134). That applies not only when such effects occur in the market on which the dominant position is held, but also when they occur in another market. It is true that an improvement in a service does not ‘immunise’ an undertaking against a finding of abuse of a dominant position, but, in the present case, not having identified any anticompetitive feature of Google’s conduct in addition to that ‘leveraging’, the Commission would not have been entitled to classify that conduct as abusive.

145

Google, supported by CCIA, states in that regard that ‘leveraging’ is an ‘umbrella’ term covering different types of abuse. For each individual type of ‘leveraging abuse’, the case-law identifies specific features of the behaviour in question that depart from competition on the merits and render that behaviour abusive, such as quality degradation, margin squeezing or a refusal to supply an indispensable input. Thus, a dominant undertaking’s practice of setting low prices cannot, by itself, be considered abusive. It is only if an additional feature that deviates from competition on the merits is identified that the practice could be characterised as predatory pricing. Thus, according to CCIA, the lack of a theoretical basis for the abuse of favouring identified by the Commission makes it impossible to ascertain the additional factors or legal principles that render such – perfectly natural – favouring an infringement of Article 102 TFEU, which creates a problem in terms of legal certainty for the internet sector as a whole.

146

The Commission, supported in that regard by the Federal Republic of Germany, states that an improvement to a service does not preclude that improvement from constituting an abuse of a dominant position, in particular if it results in a dominant undertaking favouring its own service by recourse to methods other than competition on the merits and if that is liable to produce anticompetitive effects.

147

The Commission makes clear that it disputes, moreover, that there was any improvement in Google’s general search service. While it may indeed be possible for Google to improve its general search service by showing ‘some’ grouped results on its general results pages, it would not have been able to improve its general search service by showing ‘only’ grouped results from its own comparison shopping service on its general results pages. Furthermore, the Commission recalls that, in its view, Google’s conduct cannot be justified by any objective reason related to the improvement in the quality of Google’s general search service.

148

The Federal Republic of Germany contends that Google’s impugned conduct is not within the scope of competition on the merits since it prevents competition based on the quality of the algorithm used to carry out specialised product searches. The quality of the specialised search algorithm is the constant against which the relevant undertakings compete. By means of the conduct at issue, Google encourages users to click not on the most relevant results, but on the most visible results, namely its own, irrespective of their actual relevance to the user.

149

According to VDZ, whether Google improved its service is irrelevant. The only relevant issue is whether Google used the new features of its services (Product Universals, Shopping Units, adjustment algorithms) as a vehicle to promote its own comparison shopping service at the expense of competing comparison shopping services. The improvements in Google’s comparison shopping service could, at most, be assessed from the aspect of efficiency gains. However, Google did not adduce any evidence of such efficiency gains, as required by the case-law. VDZ adds that this case is a typical example of leveraging abuse. In essence, the practices deviated from competition on the merits, because Google’s conduct on the primary market could have no economic rationale other than to foreclose competition on the secondary market. Google’s conduct in tending to favour its own comparison shopping service at the expense of competing comparison shopping services leads to the exclusion of more relevant specialised search results from competitors, which makes no economic sense.

150

According to settled case-law, a dominant undertaking has a special responsibility not to allow its behaviour to impair genuine, undistorted competition on the internal market (see judgment of 6 September 2017, Intel v Commission, C‑413/14 P, EU:C:2017:632, paragraph 135 and the case-law cited).

151

In that regard, Article 102 TFEU applies, in particular, to the conduct of a dominant undertaking that, through recourse to methods different from those governing normal competition on the basis of the performance of commercial operators, has the effect, to the detriment of consumers, of hindering the maintenance of the degree of competition existing in the market or the growth of that competition (see judgment of 27 March 2012, Post Danmark, C‑209/10, EU:C:2012:172, paragraph 24 and the case-law cited).

152

Thus, Article 102 TFEU prohibits a dominant undertaking from, among other things, adopting practices that have an exclusionary effect by using methods other than those that are part of competition on the merits (see, to that effect, judgments of 27 March 2012, Post Danmark, C‑209/10, EU:C:2012:172, paragraph 25 and the case-law cited, and of 6 September 2017, Intel v Commission, C‑413/14 P, EU:C:2017:632, paragraph 136).

153

Article 102 TFEU covers not only those practices that directly cause harm to consumers but also practices that cause consumers harm through their impact on competition (judgment of 27 March 2012, Post Danmark, C‑209/10, EU:C:2012:172, paragraph 20 and the case-law cited; see also, to that effect, judgment of 29 March 2012, Telefónica and Telefónica de España v Commission, T‑336/07, EU:T:2012:172, paragraph 171).

154

The list of abusive practices contained in Article 102 TFEU is not exhaustive, so that the list of abusive practices contained in that provision does not exhaust the methods of abusing a dominant position prohibited by EU law (judgments of 21 February 1973, Europemballage and Continental Can v Commission, 6/72, EU:C:1973:22, paragraph 26; of 17 February 2011, TeliaSonera Sverige, C‑52/09, EU:C:2011:83, paragraph 26; and of 29 March 2012, Telefónica and Telefónica de España v Commission, T‑336/07, EU:T:2012:172, paragraph 174).

155

The abuse may take the form of an unjustified difference in treatment (see, to that effect, judgments of 17 July 1997, GT-Link, C‑242/95, EU:C:1997:376, paragraph 41; of 24 October 2002, Aéroports de Paris v Commission, C‑82/01 P, EU:C:2002:617, paragraph 114; and of 7 October 1999, Irish Sugar v Commission, T‑228/97, EU:T:1999:246, paragraph 140). In that regard, the general principle of equal treatment, as a general principle of EU law, requires that comparable situations must not be treated differently and different situations must not be treated in the same way unless such treatment is objectively justified (see judgment of 16 December 2008, Arcelor Atlantique et Lorraine and Others, C‑127/07, EU:C:2008:728, paragraph 23 and the case-law cited).

156

It is, however, in no way the purpose of Article 102 TFEU to prevent an undertaking from acquiring, on its own merits, a dominant position on a market (see judgment of 6 September 2017, Intel v Commission, C‑413/14 P, EU:C:2017:632, paragraph 133 and the case-law cited).

157

Thus, not every exclusionary effect is necessarily detrimental to competition. Competition on the merits may, by definition, lead to the departure from the market or the marginalisation of competitors that are less attractive to consumers from the point of view of, among other things, price, choice, quality or innovation (see judgment of 6 September 2017, Intel v Commission, C‑413/14 P, EU:C:2017:632, paragraph 134 and the case-law cited).

158

In essence, Google maintains that the practices at issue in the contested decision are part of quality improvements in its search services and, consequently, are part of competition on the merits. Google observes, in that respect, that the Commission does not identify anything in the practices at issue that deviates from competition on the merits. Never have quality improvements in a product or service been considered by the Courts of the European Union to hinder competition.

159

As regards the Commission’s alleged failure to identify features that distinguish the practices at issue from normal competition, which was said to have been restricted, it should be noted that an undertaking’s dominant position alone, even one on the scale of Google’s position in general search services, cannot be declared unlawful under Article 102 TFEU.

160

It is apparent from settled case-law that a finding that an undertaking has a dominant position is not in itself a ground of criticism of the undertaking concerned (see judgment of 27 March 2012, Post Danmark, C‑209/10, EU:C:2012:172, paragraph 21 and the case-law cited). It is the ‘abuse’ of a dominant position that is prohibited by Article 102 TFEU.

161

It is for the Commission, in order to characterise such ‘abuse’, to identify how, by using its dominant position, the undertaking concerned has had recourse to methods different from those governing normal competition (see paragraph 151 above).

162

The mere extension of an undertaking’s dominant position to a neighbouring market cannot in itself constitute proof of conduct that departs from normal competition, even if that extension leads to the disappearance or marginalisation of competitors (see, to that effect, judgments of 27 March 2012, Post Danmark, C‑209/10, EU:C:2012:172, paragraph 22, and of 6 September 2017, Intel v Commission, C‑413/14 P, EU:C:2017:632, paragraph 134).

163

In addition, as is apparent in essence from the judgment of 25 October 2002, Tetra Laval v Commission (T‑5/02, EU:T:2002:264, paragraphs 156, 158 and 217), leveraging is a generic term in relation to the impact which a practice identified on one market may have on another market. The term may designate several different practices that are capable of being abusive, such as, in particular, tied sales as in the case giving rise to the judgment of 17 September 2007, Microsoft v Commission (T‑201/04, EU:T:2007:289), margin squeeze practices as in the case giving rise to the judgment of 29 March 2012, Telefónica and Telefónica de España v Commission (T‑336/07, EU:T:2012:172), or loyalty rebates as in the case giving rise to the judgment of 30 September 2003, Michelin v Commission (T‑203/01, EU:T:2003:250).

164

It must be noted that, while the leveraging practices of a dominant undertaking are not prohibited as such by Article 102 TFEU, the fact remains that that article is applicable to such practices. Thus, although there is no need at this stage to rule on the conditions for their prohibition, it must be noted, as indicated in paragraph 163 above, that several kinds of leveraging have previously been found to be contrary to Article 102 TFEU. In particular, in the judgment of 17 September 2007, Microsoft v Commission (T‑201/04, EU:T:2007:289, paragraph 1344), the Court considered that the practices at issue, namely bundling and the refusal to supply interoperability information, formed part of a leveraging infringement, consisting in Microsoft’s use of its dominant position on the client personal computer (PC) operating systems market to extend that dominant position to two adjacent markets.

165

Moreover, the actual scope of the special responsibility imposed on a dominant undertaking must be considered in the light of the specific circumstances of each case which show that competition has been weakened (see, to that effect, judgment of 17 February 2011, TeliaSonera Sverige, C‑52/09, EU:C:2011:83, paragraph 84 and the case-law cited).

166

In the present case, as is apparent from the contested decision and as the Commission pertinently recalled at the hearing, it did not rely solely on leveraging practices in order to conclude that there was an infringement of Article 102 TFEU.

167

The Commission considered that, through leveraging, Google was relying on its dominant position on the market for general search services in order to favour its own comparison shopping service on the market for specialised comparison shopping search services by promoting the positioning and display of that comparison shopping service and of its results on its general results pages, as compared to competing comparison shopping services, whose results, given their inherent characteristics, were prone to being demoted on those pages by adjustment algorithms.

168

The Commission observed in that regard, in recital 344 of the contested decision, that while results from competing comparison shopping services could appear only as generic results, that is to say, simple blue links that were also prone to being demoted by adjustment algorithms, results from Google’s own comparison shopping service were prominently positioned at the top of Google’s general results pages, displayed in rich format and incapable of being demoted by those algorithms, resulting in a difference in treatment in the form of Google’s favouring of its own comparison shopping service.

169

The Commission explained in particular that, on account of three specific circumstances – namely (i) the importance of traffic generated by Google’s general search engine for comparison shopping services (Section 7.2.2 of the contested decision); (ii) the behaviour of users when searching online (Section 7.2.3 of the contested decision); and (iii) the fact that diverted traffic from Google’s general results pages accounts for a large proportion of traffic to competing comparison shopping services and cannot be effectively replaced by other sources (Section 7.2.4 of the contested decision) – this favouring was liable to lead to a weakening of competition on the market.

170

In the first place, as regards the importance of traffic generated by Google’s general search engine, in Section 7.2.2 of the contested decision (recitals 444 to 450), the Commission pointed out that this was, according to the statement of a competing comparison shopping service set out in recital 444 of the contested decision, the ‘most important “asset” of a [specialised] search engine’. The Commission thus explained that that traffic increased the relevance of specialised search results and in particular the freshness and breadth of the offering of comparison shopping services by enhancing their ability to convince merchants to provide them with data about their products (recital 445), that it generated revenue either via commissions paid by merchants or online advertising (recital 446), and that it provided information about user behaviour, which improved the relevance and usefulness of results, including through machine learning effects (recital 447), experiments (recital 448) or the suggestion of other search terms that might be of interest for users (recital 449).

171

The Commission thus stated, in essence, in Section 7.2.2 of the contested decision, that that traffic produced network effects, in that the more a comparison shopping service is visited by internet users, the greater the relevance and usefulness of its services and the more merchants would be inclined to use them, and that that traffic also generated revenue from commissions or advertising that could be used to improve the usefulness of the services provided and thus distinguish that comparison shopping service from competitors. In other words, the Commission explained that generating traffic initiated a virtuous circle, improving the relevance of results and thus attracting more users and ultimately more revenue from advertising partners or online sellers who placed their products on the website of the comparison shopping service, which in turn meant that the undertaking concerned could invest more in improving or, at the very least, maintaining its competitive position in a sector – the digital sector – in which innovation is key to commercial success. Conversely, loss of traffic can lead to a vicious circle and, eventually, to market exit due to an inability to compete on essential elements such as the relevance of results and innovation, which are linked, since comparison shopping services innovate in order to improve the relevance of their results and thus attract more traffic and therefore more revenue.

172

In the second place, as regards user behaviour, the Commission indicated that the favouring in which Google engaged, its own results being displayed more visibly and those of its competitors less so, was capable of influencing the behaviour of internet users when they wished to consult comparison shopping websites (Section 7.2.3.1 and recitals 454 to 461 of the contested decision). The Commission explained in that regard, in recitals 455 to 457 of the contested decision, that users typically concentrated on the first three to five search results and paid little or no attention to the remaining results, particularly those below the part of the screen that was immediately visible (the fold). The Commission thus argued, in recital 535 of the contested decision, that users tended to assume that the most visible results were the most relevant, irrespective of their actual relevance.

173

In the third place, as regards the impact of diverted traffic, the Commission stated that this accounted for a large proportion of traffic to competing comparison shopping services (Section 7.2.4.1 of the contested decision) and could not be effectively replaced by other sources, including text ads, mobile applications (apps), direct traffic, referrals from affiliate websites, social networks or other search engines (Section 7.2.4.2 of the contested decision).

174

The importance of Google traffic from its general search pages and the nature of that traffic as being not effectively replaceable were, in view of the background recalled in paragraphs 168 to 173 above and without any error of law being committed, treated by the Commission as relevant circumstances capable of characterising the existence of practices falling outside the scope of competition on the merits.

175

It follows from the foregoing that the Commission did not merely identify leveraging but, as required by the case-law, classified Google’s accompanying practices in law on the basis of relevant criteria. Thus, if the favouring and its effects, identified in the light of the specific circumstances of the relevant markets, have been validly demonstrated by the Commission, which will be verified having regard to all of the pleas and arguments, the Commission will have been fully entitled to take the view that that favouring was a departure from competition on the merits.

176

It must be observed in that regard that, given the universal vocation of Google’s general search engine, which, as is apparent from recital 12 of the contested decision, is designed to index results containing any possible content, the promotion on Google’s general results pages of one type of specialised result – its own – over the specialised results of competitors involves a certain form of abnormality.

177

The infrastructure at issue, namely Google’s general results pages which generate traffic to other websites, including those of competing comparison shopping services, is, in principle, open, which distinguishes it from other infrastructures referred to in the case-law, consisting of tangible or intangible assets (press distribution systems or intellectual property rights, respectively) whose value depends on the proprietor’s ability to retain exclusive use of them.

178

Unlike the latter infrastructures, the rationale and value of a general search engine lie in its capacity to be open to results from external (third-party) sources and to display these multiple and diverse sources on its general results pages, sources which enrich and enhance the credibility of the search engine as far as the general public is concerned, and enable it to benefit from the network effects and economies of scale that are essential for its development and its subsistence in a market in which, by their very nature, few infrastructures of that kind can subsist, given those network effects. A very large number of users is needed to reach the critical mass capable of compensating for the service being free of charge on one side of the market and generating advertising income on its other side. Accordingly, for a search engine, limiting the scope of its results to its own entails an element of risk and is not necessarily rational, save in a situation, as in the present case, where the dominance and barriers to entry are such that no market entry within a sufficiently short period of time is possible in response to that limitation of internet users’ choice.

179

Consequently, the fact, assuming it to be established, that Google favours its own specialised results over third-party results, which seems to be the converse of the economic model underpinning the initial success of its search engine, cannot but involve a certain form of abnormality. It follows that, in accordance with the case-law cited in paragraph 133 above, it is for the person responsible for that difference in treatment to justify it in the light of competition law (see, to that effect, judgment of 12 December 2018, Servier and Others v Commission, T‑691/14, under appeal, EU:T:2018:922, paragraph 1377 and the case-law cited).

180

It may be observed moreover, for the sake of completeness, that even in a situation that differs from that of the present case, the Court of Justice has ruled, with regard to internet access providers, that the EU legislature had intended, by Regulation (EU) 2015/2120 of the European Parliament and of the Council of 25 November 2015 laying down measures concerning open internet access and amending Directive 2002/22/EC on universal service and users’ rights relating to electronic communications networks and services and Regulation (EU) No 531/2012 on roaming on public mobile communications networks within the Union (OJ 2015 L 310, p. 1), to impose on those operators a general obligation of equal treatment, without discrimination, restriction or interference with traffic, from which derogation is not possible in any circumstances by means of commercial practices (see, to that effect, judgment of 15 September 2020, Telenor Magyarország, C‑807/18 and C‑39/19, EU:C:2020:708, paragraph 47). The fact that the legislature made that choice and the legal obligation of non-discrimination that follows from it for internet access providers on the upstream market cannot be disregarded when analysing the practices of an operator like Google on the downstream market, given the undisputed ultra-dominant position of Google on the market for general search services and its special responsibility not to allow its behaviour to impair genuine, undistorted competition in the internal market. It is of no relevance in that regard whether or not legislation calls, in general terms, for such non-discriminatory access to online search results, since, as is clear from the case-law, a system of undistorted competition can be guaranteed only if equality of opportunity is secured as between the various economic operators (see judgment of 14 October 2010, Deutsche Telekom v Commission, C‑280/08 P, EU:C:2010:603, paragraph 230 and the case-law cited), which is consistent with the possibility that certain differences in treatment may be considered contrary to Article 102 TFEU when what is at issue are favouring practices established by operators in a dominant position in the internet sector.

181

Furthermore, as VDZ submits, the deviation in relation to competition on the merits of the conduct at issue, assuming it is established, is all the more obvious as it follows a change of conduct on the part of the dominant operator. Google did indeed change its conduct on the market for general search services.

182

It is apparent from the file that, historically, Google initially provided general search services and acquired a ‘superdominant’ position on that market, which is characterised by very high barriers to entry. On that market, Google displayed results that directed users to comparison shopping services. Furthermore, Google displayed all the results of specialised search services in the same way and according to the same criteria. The very purpose of a general search service is to browse and index the greatest possible number of web pages in order to display all results corresponding to a search.

183

Google subsequently entered the market for specialised comparison shopping search services. At the time when Google started its activities on the market for specialised comparison shopping search services, there were already numerous providers of such services. Moreover, in view of its ‘superdominant’ position, its role as a gateway to the internet and the very high barriers to entry on the market for general search services, it was under a stronger obligation not to allow its behaviour to impair genuine, undistorted competition on the related market for specialised comparison shopping search services.

184

According to the Commission, after Google’s launch on the market for specialised comparison shopping search services and after experiencing the failure of its dedicated comparison shopping web page (Froogle), Google changed its practices on the market for general search services which it dominated, the effect of which was to increase the visibility of results from its own comparison shopping service on the general search results pages. After the launch of grouped product results, comparison shopping services were no longer all treated in the same way. Google promoted its own specialised search results (positioning and display) and demoted the results of its competitors which, moreover, were not afforded the same type of display (only ‘blue links’ without images or rich text). The change in Google’s behaviour led to a reduction in the visibility of results from competing comparison shopping services and, at the same time, increased the visibility of results from Google’s own comparison shopping service. Thus, the practices at issue enabled Google to highlight its own comparison shopping service on its general search results pages while leaving competing comparison shopping services virtually invisible on those pages, which, in principle, is not consistent with the intended purpose of a general search service.

185

Accordingly, subject to the favouring and its effects identified at the end of the analysis summarised in paragraphs 170 to 173 above having been properly established, Google’s conduct cannot, as such, constitute competition on the merits.

186

That conclusion is not undermined by Google’s arguments to the effect that the display of Product Universals and Shopping Units cannot be classified as abusive, since those results and those ads constituted quality improvements in its services that were within the scope of competition on the merits.

187

First, it should be pointed out that Google’s arguments are based on the incorrect premiss that the conduct at issue consists solely in the special display and positioning of Product Universals and Shopping Units, when in fact that conduct consists in the combination of two practices: the promotion of specialised results from Google’s comparison shopping service and the simultaneous demotion of results from competing comparison services by adjustment algorithms. It must be noted in that regard that Google does not describe the demotion on its general results pages of competing comparison shopping services, but not of its own, as a ‘quality improvement’ that would characterise competition on the merits.

188

Secondly, contrary to what is suggested by Google, it does not follow from any of the judgments cited by the Commission in recital 334 of the contested decision that conduct leading to a product or service improvement cannot constitute, in itself, an autonomous form of abuse where that improvement results in the dominant undertaking favouring its own product or service through recourse to methods different from those governing competition on the merits and that conduct is capable of having anticompetitive effects. In that regard, as VDZ correctly points out, product or service improvements of a technical or commercial nature can be taken into account only at the stage when any objective justifications and possible efficiency gains that might thereby be achieved are being examined.

189

The conclusion that Google’s conduct, if it is established that it qualifies as favouring, may depart from competition on the merits is not undermined by CCIA’s arguments that the lack of a clear legal test in the contested decision breaches the principle of legal certainty.

190

It must be stated at the outset that the Commission disputes the admissibility of that argument, maintaining that it is inadmissible because, in essence, it is new as compared to Google’s arguments.

191

In accordance with the fourth paragraph of Article 40 of the Statute of the Court of Justice of the European Union, applicable to the proceedings before the General Court by virtue of the first paragraph of Article 53 of that statute, an application to intervene is to be limited to supporting the form of order sought by one of the parties. Thus, a party which is granted leave to intervene in a case submitted to the General Court may not alter the subject matter of the dispute as defined by the forms of order sought by the main parties and the pleas in law raised by those parties. It follows that arguments submitted by an intervener are admissible only if they come within the framework provided by those forms of order and pleas in law (judgment of 29 July 2019, Bayerische Motoren Werke and Freistaat Sachsen v Commission, C‑654/17 P, EU:C:2019:634, paragraph 50). In addition, under Article 142(3) of the Rules of Procedure, the intervener must accept the case as he or she finds it at the time of the intervention.

192

In this respect, it should be pointed out that while those provisions do not preclude the intervener from advancing arguments which are new or which differ from those of the party he or she supports, lest the intervention be limited to restating the arguments advanced in the application, it cannot be held that those provisions permit that intervener to alter or distort the context of the dispute defined in the application by raising new pleas in law (judgment of 12 December 2006, SELEX Sistemi Integrati v Commission, T‑155/04, EU:T:2006:387, paragraph 42).

193

Nevertheless, in the present case, the argument at issue is put forward in support of Google’s contention that, in a reversal of the case-law relating to leveraging abuses, the Commission has not identified in this particular case any specific feature that distinguishes the conduct at issue from conduct constituting competition on the merits, and draws a specific conclusion from this for the members of CCIA, namely breach of the principle of legal certainty. In those circumstances, the argument is admissible.

194

As regards the assessment of the merits of that argument, it must be recalled that observance of the principle of legal certainty requires that the institutions avoid, as a matter of principle, inconsistencies that might arise in the implementation of the various provisions of EU law (see judgment of 22 April 2016, Ireland and Aughinish Alumina v Commission, T‑50/06 RENV II and T‑69/06 RENV II, EU:T:2016:227, paragraph 59 and the case-law cited).

195

In the present case, recital 341 of the contested decision sets out the reasons why the practices at issue depart from competition on the merits, stating, in essence, that they diverted traffic and, moreover, that they are capable of having anticompetitive effects. Accordingly, by that recital, read in isolation, the Commission seems to have inferred from the existence of exclusionary effects arising from those practices that they deviate from competition on the merits. Such a description relating solely to the exclusionary effects of the practices could give rise to queries as to whether the test which the Commission used to characterise the infringement of Article 102 TFEU is consistent with the principle of legal certainty. It follows from the case-law cited in paragraph 157 above that any practice, whether a pricing practice or not, which has exclusionary effects cannot be regarded, on that basis alone, as being anticompetitive.

196

However, recital 341 of the contested decision must be read in conjunction with recital 342 of that decision, in which the Commission states, ‘to demonstrate why the Conduct is abusive and falls outside the scope of competition on the merits’, that the practices at issue consist in Google favouring its own comparison shopping service at the expense of competing comparison shopping services and that that favouring occurs within a particular context. In that recital the Commission lists the numerous aspects which it took into account to demonstrate why the practice is abusive and deviates from competition on the merits and, in particular, as is apparent from paragraphs 170 to 173 above, the three criteria relating to the importance of traffic generated by Google’s general search engine for comparison shopping services (Section 7.2.2 of the contested decision), user behaviour when carrying out online searches (Section 7.2.3.1 of the contested decision) and the fact that the traffic diverted cannot be effectively replaced (Section 7.2.4 of the contested decision).

197

Thus, the Commission’s analysis resulting in a finding of abuse is not in any way ‘inconsistent’, within the meaning of the case-law cited in paragraph 194 above, with the case-law on abusive leveraging cited in recital 334 of the contested decision, in so far as it may be concluded that there is an infringement on the basis, first, of suspect elements in the light of competition law (in particular an unjustified difference in treatment) which are absent in the case of a refusal of access and, secondly, of specific circumstances, in accordance with the case-law referred to in paragraph 165 above, relating to the nature of the infrastructure from which that difference in treatment arises (in this instance, importance and being not effectively replaceable, in particular).

198

In those circumstances, the first part of the fifth plea in law must be rejected.

199

The second part of the fifth plea for annulment seeks a finding from the Court that the Commission was not entitled to require Google to give competing comparison shopping services access to the services resulting from its comparison shopping improvements without satisfying the conditions identified in the case-law and, in particular, those applicable to infrastructures qualifying as essential facilities.

200

First, Google states that that is indeed the import of the contested decision, which imposes on it a duty to supply, even though the conduct at issue is described only as favouring, in the sense that Google treated its own search results more favourably than those of its competitors. Google relies in that regard, in particular, on recitals 538 and 662 of the contested decision, the latter stating that ‘the abuse established by this Decision concerns simply the fact that Google does not position and display in the same way results from Google’s comparison shopping service and from competing comparison shopping services’. Google asserts that the contested decision identifies no criteria or principles that distinguish the infringement at issue from a duty to supply case. It is irrelevant that the decision used a different form of words to punish a refusal to supply. The need to apply the criteria relating to a duty to supply depends on the substance and the nature of that obligation, not on the way in which it is worded.

201

In Google’s submission, the favouring claim in the contested decision really concerns the access of competing comparison shopping services to Google’s ‘technologies and designs’, in that the Commission does not seek to prevent Google from showing Product Universals and Shopping Units (recitals 656 and 662 of the contested decision). Instead, it objects that Google does not position and display competing comparison shopping services in the same way, which would entail their having access to those ‘technologies and designs’. The same argument as that raised by the Commission in the contested decision in order to find that there was favouring could have been raised in the case giving rise to the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), since the press publisher concerned, Mediaprint, included its newspapers in its distribution network, but not those of its competitor. Similarly, in the case giving rise to the judgment of 3 October 1985, CBEM (311/84, EU:C:1985:394), a claim could have been made that the television station concerned favoured its own telemarketing services by only allowing advertisements that included its own telephone number. Thus, according to Google, if the contested decision were to be upheld, any duty to supply could be re-characterised as an act of favouring, without any need to meet the indispensability condition established by the Court of Justice in its case-law. All the judgments in which it required that condition to be met would be undermined. As the Bundesgerichtshof (Federal Court of Justice, Germany) stated, when seised of complaints concerning similar favouring cases, undertakings are not required to subsidise their competitors.

202

Furthermore, it is not claimed that Google has created barriers to entry or restrictions preventing competing comparison shopping services from attracting traffic from third-party sources. If the alleged anticompetitive effects were derived from a lack of access to Google’s traffic, it was for the Commission to demonstrate, in accordance with the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), that such access was ‘indispensable’ for competition and that lack of access risked eliminating all competition.

203

Secondly, Google submits that the contested decision does not demonstrate that access to its services was indispensable for competing comparison shopping services and that, without such access, all effective competition could be eliminated, conditions which are necessary, according to the case-law, for a duty to supply to be imposed on a dominant undertaking. The contested decision thus states only that Google search traffic is ‘important for the ability of a comparison shopping service to compete’ (recital 444) without ever demonstrating that that traffic is ‘indispensable’, and merely asserts that other traffic sources are less effective for competing comparison shopping services (recital 542).

204

Thirdly, Google adds that, in the contested decision, the Commission wrongly departed from the case-law on the duty to supply by putting forward two reasons that are incorrect. First of all, in recital 650 of the contested decision, it stated that Google’s conduct did not consist simply in a passive refusal to give access to its general results pages, but in active behaviour favouring its own comparison shopping service by more favourable positioning and display in those pages. According to Google, in the case giving rise to the judgment of 3 October 1985, CBEM (311/84, EU:C:1985:394, paragraph 5), to cite one example, although the conduct at issue was also active, the Court of Justice drew attention to the indispensability of the service that had been refused and the risk of eliminating all competition and found that a dominant undertaking could not reserve that service to itself.

205

Next, in recital 651 of the contested decision, the Commission found that the indispensability criterion did not apply, as the decision required only that ‘Google … cease the Conduct’ and did not require Google to transfer an asset or enter into new agreements. However, even though it was open to Google to cease using the services at issue for its own benefit, instead of giving access to them by entering into agreements with interested parties, that same choice was also available to undertakings that had been made subject to a duty to supply in order to bring an end to an abuse of a dominant position.

206

In short, by the contested decision, the Commission objected to the improvements relating to search results and product ads and their underlying technologies because Google had not provided competing comparison shopping services with access to them. In order to make a finding of abuse on the basis of that reasoning, the Commission would have had to show that such access was indispensable and that lack of access risked eliminating all competition.

207

CCIA submits in more general terms that the contested decision is based on the false premiss that Google’s search engine is a gateway to the internet. Today, more than ever, there are numerous points of entry for competing online and no one site is a gateway to the internet.

208

The Commission, supported by the Federal Republic of Germany, VDZ, Twenga and Kelkoo, maintains that the criteria set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), are not applicable in this case. It reiterates the arguments set out in the contested decision, referred to in paragraphs 204 and 205 above, and maintains that it left it to Google to decide how to ensure equal treatment of its own comparison shopping service and competing comparison shopping services, which covered either the possibility of continuing to display Shopping Units on its general results pages by incorporating, by contract, results from competing comparison shopping services, or the possibility of no longer displaying Shopping Units on that page.

209

The Commission disputes Google’s argument that abuse of a dominant position may be established only if the conditions relating to a refusal to supply an ‘essential facility’ are satisfied, while other anticompetitive conduct with the effect of extending or strengthening a dominant position on a market may exist. As long as the Commission demonstrates that competition may be restricted by the anticompetitive conduct of a dominant undertaking, it is not obliged to demonstrate that there was a refusal on the part of that undertaking to supply a product or service that is indispensable for its competitors. The Commission cites the example of the case giving rise to the judgment of 23 October 2003, Van den Bergh Foods v Commission (T‑65/98, EU:T:2003:281, paragraphs 159 and 161).

210

The Federal Republic of Germany contends in support of the Commission that, unlike the case at issue in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), access to an ‘essential facility’ is not at issue in the present case. It claims that Google already ‘supplies’ its competitors by giving them access to its general search service. As in the situation giving rise to that judgment, there is no exclusion of competitors. On the contrary, the Commission takes issue with Google for displaying competitors’ services less favourably than its own service, since competitors’ results are shown in a way that suggests they are less relevant than Google’s results.

211

VDZ asserts that the conduct at issue is a typical example of leveraging abuse comparable to practices that have already been found to be unlawful, like bundling and tying, margin squeezes and particular types of refusal to deal, and that this conduct was treated as such.

212

In the contested decision, the Commission concluded that the conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569) were not applicable to the facts of the present case for three reasons in particular. The Commission found, in the first place, that abusive leveraging constituted a well-established, independent form of abuse falling outside the scope of competition on the merits (recital 649); in the second place, that the practices at issue did not concern a passive refusal of access to Google’s general results pages, but active favouring in the form of Google’s promotion of its own comparison shopping service compared to competing comparison shopping services (recital 650); and, in the third place, that it was not necessary in this case that the undertaking at issue transfer an asset or enter into agreements with persons with whom it had not chosen to contract, in order to bring the abuse to an end. The Commission justified its assertion that the judgment of 26 November 1998, Bronner (C‑7/97,EU:C:1998:569) was not applicable on the basis of the third reason by citing the judgment of 23 October 2003, Van den Bergh Foods v Commission (T‑65/98, EU:T:2003:281, paragraph 161) (recital 651).

213

As a preliminary point, it should be recalled that, in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), the Court of Justice considered that, in order for the refusal by an undertaking in a dominant position to grant access to a service to be capable of constituting an abuse within the meaning of Article 102 TFEU, it was necessary that that refusal be likely to eliminate all competition in the market on the part of the person requesting the service, that such refusal be incapable of being objectively justified and that the service in itself be indispensable to carrying on that person’s business, inasmuch as there was no actual or potential substitute for it (judgment of 26 November 1998, Bronner, C‑7/97, EU:C:1998:569, paragraph 41; see also judgment of 9 September 2009, Clearstream v Commission, T‑301/04, EU:T:2009:317, paragraph 147 and the case-law cited).

214

In essence, in the context of the second part of the fifth plea, Google claims that the Commission treated the practices at issue as a ‘refusal to supply’ without verifying, in particular, that access to the elements concerned, namely, the general results pages and its own specialised results (Product Universals and Shopping Units), was ‘indispensable’ and that there was a risk of all competition being eliminated, as it ought to have done in the light of the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569). According to Google, the Commission thus penalised a refusal to supply while exempting itself from the conditions and evidential burden of establishing that infringement.

215

The conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569) apply, in principle, to infrastructures or to services that are often described as an ‘essential facility’ in the sense that they are indispensable for carrying on a business on a market where there is no actual or potential substitute (see judgments of 15 September 1998, European Night Services and Others v Commission, T‑374/94, T‑375/94, T‑384/94 and T‑388/94, EU:T:1998:198, paragraphs 208 and 212 and the case-law cited, and of 9 September 2009, Clearstream v Commission, T‑301/04, EU:T:2009:317, paragraph 147 and the case-law cited), so that refusing access may lead to the elimination of all competition. The case-law on essential facilities relates in particular to situations in which the free exercise of an exclusive right, being a right which rewards investment or innovation, may be limited in the interest of undistorted competition in the internal market (see judgments of 1 July 2010, AstraZeneca v Commission (T‑321/05, EU:T:2010:266, paragraph 679, and of 18 November 2020, Lietuvos geležinkeliai v Commission, T‑814/17, under appeal, EU:T:2020:545, paragraph 87 and the case-law cited).

216

On numerous occasions (judgments of 6 March 1974, Istituto Chemioterapico Italiano and Commercial Solvents v Commission, 6/73 and 7/73, EU:C:1974:18, paragraph 25; of 3 October 1985, CBEM, 311/84, EU:C:1985:394, paragraph 26; of 6 April 1995, RTE and ITP v Commission, C‑241/91 P and C‑242/91 P, EU:C:1995:98, paragraph 56; of 26 November 1998, Bronner, C‑7/97, EU:C:1998:569, paragraph 41; of 29 April 2004, IMS Health, C‑418/01, EU:C:2004:257, paragraph 52; of 12 June 1997, Tiercé Ladbroke v Commission, T‑504/93, EU:T:1997:84, paragraph 132; and of 17 September 2007, Microsoft v Commission, T‑201/04, EU:T:2007:289, paragraph 332), the Courts of the European Union, guided by the doctrine of essential facilities, have used the criteria of indispensability and of the risk of eliminating all competition to characterise or to rule out the existence of an abuse in cases concerning the possibility of a dominant undertaking reserving to itself an activity on a neighbouring market.

217

As Advocate General Jacobs explains in essence in his Opinion in Bronner (C‑7/97, EU:C:1998:264, points 56, 57 and 62), the choice of the criterion of indispensability, and that relating to the risk of eliminating all competition, reflect the desire, from a legal point of view, to protect the right of an undertaking to choose its trading partners and freely to dispose of its property, principles that are generally recognised in the laws of the Member States, in some cases with constitutional status, and, from an economic point of view, in the long term, to promote competition, in the interest of consumers, by allowing a company to retain for its own use facilities which it has developed. The purpose of the three conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), and recalled in paragraph 213 above, is thus to ensure that a duty imposed on an undertaking in a dominant position to provide access to its infrastructure does not ultimately impede competition by reducing that undertaking’s initial incentive to build such infrastructure. Indeed, the incentive for a dominant undertaking to invest in facilities would be reduced if its competitors were, upon request, able to share the benefits (judgment of 18 November 2020, Lietuvosgeležinkeliai v Commission, T‑814/17, under appeal, EU:T:2020:545, paragraph 90).

218

It is in the light of those preliminary considerations that the Court must examine the arguments raised by Google to the effect that the Commission failed to have regard to Article 102 TFEU by penalising the practices at issue without establishing that the conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569) were satisfied, in particular, the condition of indispensability.

219

In the first place, contrary to the Commission’s contention, what is at issue in the present case are the conditions of the supply by Google of its general search service by means of access to general results pages for competing comparison shopping services, such access being, as is apparent from Section 7.2.2 of the contested decision, presented as ‘important’ for generating traffic on comparison shopping services’ websites and therefore ultimately revenue and, as is apparent from Section 7.2.4.2, ‘not effectively replaceable’.

220

Thus, as is apparent from recital 662 of the contested decision, Google is accused of failing to make a similar type of positioning and display available to competing comparison shopping services as is available to its own comparison service, and therefore of failing to ensure equal treatment of its own comparison service and the services of its competitors.

221

The contested decision states in that regard, in recital 699, that any ‘measure’ of implementation must ensure that Google does not treat competing comparison shopping services ‘less favourably’ than its own comparison shopping service within its general results pages and, in recital 700(c), that any such measure should subject Google’s own comparison shopping service to the ‘same … processes and methods’ for positioning and display as those used for competing comparison shopping services.

222

The contested decision thus envisages equal access by Google’s comparison shopping service and competing comparison shopping services to Google’s general results pages, irrespective of the type of result concerned (generic results, Product Universals or Shopping Units), and does therefore seek to provide competing comparison shopping services with access to Google’s general results pages and ensure that their positioning and display within those pages are as visible as those of Google’s comparison shopping service, even if it does not rule out the possibility that, in order to implement the remedy required by the Commission, Google will cease to display and position its own comparison shopping service more favourably than competing comparison shopping services on its general results pages.

223

In the second place, it must be noted that, faced with that issue of access, as is apparent from recitals 649 to 652 of the contested decision, the Commission did not refer, or at least not expressly, to the conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569) in finding the abuse to have been established. On the contrary, as is apparent from recitals 334 and 649 of the contested decision, the Commission relied on the case-law applicable to abusive leveraging in order to conclude that the anticompetitive practices at issue were established. The Commission found in that regard that Google was leveraging its dominant position on the market for general search services in order to favour its own comparison shopping service on the market for comparison shopping services, such favouring leading to the potential or actual foreclosure of competition on the downstream market (recitals 341 and 342 of the contested decision).

224

It must be noted that Google’s general results page has characteristics akin to those of an essential facility (see, to that effect, judgments of 15 September 1998, European Night Services and Others v Commission, T‑374/94, T‑375/94, T‑384/94 and T‑388/94, EU:T:1998:198, paragraphs 208 and 212 and the case-law cited, and of 9 September 2009, Clearstream v Commission, T‑301/04, EU:T:2009:317, paragraph 147 and the case-law cited), inasmuch as there is currently no actual or potential substitute available that would enable it to be replaced in an economically viable manner on the market (see, to that effect, judgment of 17 September 2007, Microsoft v Commission, T‑201/04, EU:T:2007:289, paragraphs 208, 388, 390, 421 and 436).

225

It must be noted in that regard, as is apparent from paragraphs 170 to 173 above, that the Commission found, in Section 7.2.4 of the contested decision, that generic search traffic from Google’s general results pages accounted for a large proportion of traffic to competing comparison shopping services and that such traffic could not be effectively replaced by other sources of traffic currently available to comparison shopping services, factors which are presented as essential aspects in the analysis of the abusive conduct.

226

The Commission thus made clear, in Section 7.2.4.2 of the contested decision, that there was currently no viable alternative for traffic accounting for a large proportion of the activity of comparison shopping services. In recital 588 of the contested decision, the Commission stated that ‘traffic from other general search services (such as Bing or Yahoo) is insignificant and unlikely to increase due to the barriers to entry to the national markets for general search services’. In recitals 285 to 305 of the contested decision, the Commission described the barriers to entry to the markets for general search services. It found that those barriers resulted from significant investments and the effects of scale and network effects. It explained how the history of general search services markets bore out the existence of those barriers to entry, with only one significant launch on the market since 2009 (that of Microsoft with Bing) and Google’s (almost worldwide) quasi-monopoly. In recital 544 of the contested decision, the Commission also explained that increased investment in text ads to compensate for the loss of traffic from Google’s search engine was not an ‘economically viable’ solution, nor were other sources of traffic such as mobile apps or direct traffic (recitals 568 and 580).

227

In so doing, by finding that the traffic generated by Google’s general search pages was not ‘effectively replaceable’ and that other sources of traffic were not ‘economically viable’, the Commission considered Google’s traffic to be indispensable for competing comparison shopping services (see, to that effect and by analogy with regard to a computer operating system with similar characteristics, judgment of 17 September 2007, Microsoft v Commission, T‑201/04, EU:T:2007:289, paragraphs 208, 388, 390, 421 and 436).

228

Lastly, in Section 7.3 of the contested decision, the Commission concluded that the practices at issue could lead to the potential elimination of all competition. The Commission thus stated, in recital 594 of the contested decision, that those practices were ‘capable of leading competing comparison shopping services to cease providing their services’.

229

In the third place, it should be noted that while the practices at issue, as Google maintains, are not unrelated to the issue of access, they can nevertheless be distinguished in their constituent elements from the refusal to supply at issue in the case giving rise to the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), which vindicates the Commission’s decision to consider them from the aspect of criteria other than those specific to that judgment.

230

Not every issue of, or partly of, access, like that in the present case, necessarily means that the conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569) relating to the refusal to supply must be applied.

231

That is so in particular, as the Commission indicates in recital 649 of the contested decision (see paragraph 212 above), where the practice at issue consists in independent conduct which can be distinguished, in its constituent elements, from a refusal to supply, even if it may have the same exclusionary effects.

232

A ‘refusal’ to supply that warrants the application of the conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569) implies (i) that it is express, that is to say, that there is a ‘request’ or in any event a wish to be granted access and a consequential ‘refusal’, and (ii) that the trigger of the exclusionary effect – the impugned conduct – lies principally in the refusal as such, and not in an extrinsic practice such as, in particular, another form of leveraging abuse (see, to that effect, judgments of 6 March 1974, Istituto Chemioterapico Italiano and Commercial Solvents v Commission, 6/73 and 7/73, EU:C:1974:18, paragraphs 24 and 25; of 3 October 1985, CBEM, 311/84, EU:C:1985:394, paragraphs 26 and 27; of 6 April 1995, RTE and ITP v Commission, C‑241/91 P and C‑242/91 P, EU:C:1995:98, paragraphs 9, 11, 54 and 55; of 26 November 1998, Bronner, C‑7/97, EU:C:1998:569, paragraphs 8, 11 and 47; of 12 June 1997, Tiercé Ladbroke v Commission, T‑504/93, EU:T:1997:84, paragraphs 5, 7, 110, 131 and 132; and of 17 September 2007, Microsoft v Commission, T‑201/04, EU:T:2007:289, paragraphs 2 and 7).

233

Conversely, the lack of such an express refusal to supply precludes practices from being described as a refusal to supply and analysed with respect to the strict conditions laid down for such a refusal where, notwithstanding that those practices might ultimately result in an implicit refusal of access, they constitute, in view of their constituent elements which deviate, by their very nature, from competition on the merits, an independent infringement of Article 102 TFEU.

234

As confirmed, moreover, by Advocate General Saugmandsgaard Øe in his Opinion in Deutsche Telekom v Commission and Slovak Telekom v Commission (C‑152/19 P and C‑165/19 P, EU:C:2020:678, points 85 to 89), all or, at the very least, most practices capable of restricting or eliminating competition (‘exclusionary practices’) are liable to constitute implicit refusals to supply, since they tend to make access to a market more difficult. Nonetheless, the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), cannot be applied to all of those practices without disregarding the spirit and the letter of Article 102 TFEU, the scope of which is not limited to abusive practices relating to ‘indispensable’ goods and services within the meaning of that judgment.

235

It must, moreover, be observed that in a number of cases which, like the present case, raised issues of access to a service, it was not necessary to demonstrate that the condition as to indispensability was satisfied. That was so, inter alia, in relation to margin squeezing (judgments of 17 February 2011, TeliaSoneraSverige, C‑52/09, EU:C:2011:83, paragraphs 55 to 58, and of 10 July 2014, Telefónica and Telefónica de España v Commission, C‑295/12 P, EU:C:2014:2062, paragraph 75) and tied sales (judgment of 17 September 2007, Microsoft v Commission, T‑201/04, EU:T:2007:289, paragraph 961).

236

In that regard, as the Court of Justice has held, it cannot be inferred from paragraphs 48 and 49 of the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569) that the conditions to be met in order to establish that a refusal to supply is abusive must necessarily also apply when assessing the abusive nature of conduct which consists in supplying services or selling goods on conditions which are disadvantageous or on which there might be no purchaser. Such conduct may, in itself, constitute an independent form of abuse distinct from that of refusal to supply (judgment of 17 February 2011, TeliaSonera Sverige, C‑52/09, EU:C:2011:83, paragraphs 55 and 56; see also, to that effect, judgment of 10 July 2014, Telefónica and Telefónica de España v Commission, C‑295/12 P, EU:C:2014:2062, paragraphs 75 and 96).

237

It should be noted in that regard, as is apparent from Section 7.2.3 of the contested decision, that the practices at issue are based, according to the Commission, on internal discrimination between Google’s own comparison shopping service and competing comparison shopping services through leveraging from a dominated market, characterised by high barriers to entry, namely the market for general search services.

238

Consequently, as is apparent from recital 344 and Article 1 of the contested decision, the present case is not concerned merely with a unilateral refusal by Google to supply a service to competing undertakings that is necessary in order to compete on a neighbouring market, which would be contrary to Article 102 TFEU and would therefore justify the application of the ‘essential facilities’ doctrine (see, to that effect, judgment of 22 March 2011, Altstoff Recycling Austria v Commission, T‑419/03, EU:T:2011:102, paragraph 109), but with a difference in treatment that is contrary to the provisions of that article.

239

The Advocates General of the Court of Justice have consistently distinguished cases of difference in treatment from cases of refusal of access by excluding the application of the conditions derived from the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569). That exclusion was referred to by Advocate General Jacobs in his Opinion in Bronner (C‑7/97, EU:C:1998:264, point 54), and by Advocate General Mazák, who expressly ruled out the application of the indispensability condition in cases where ‘the dominant undertaking may be discriminating between competitors and its own downstream operations under Article 102(c) TFEU’ (Opinion of Advocate General Mazák in TeliaSonera Sverige, C‑52/09, EU:C:2010:483, point 32), and is confirmed by the General Court in the judgment of 7 October 1999, Irish Sugar v Commission (T‑228/97, EU:T:1999:246, paragraphs 166 and 167).

240

It must therefore be concluded that the Commission was not required to establish that the conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), were satisfied in order to make a finding of an infringement on the basis of the practices identified, since, as the Commission states in recital 649 of the contested decision, the practices at issue are an independent form of leveraging abuse which involve, as the Commission also indicates in recital 650 of that decision, ‘active’ behaviour in the form of positive acts of discrimination in the treatment of the results of Google’s comparison shopping service, which are promoted within its general results pages, and the results of competing comparison shopping services, which are prone to being demoted. They can thus be distinguished from the conduct at issue in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), which consisted in a simple refusal of access, as the Court of Justice moreover pointed out in the judgment of 25 March 2021, Deutsche Telekom v Commission (C‑152/19 P, EU:C:2021:238, paragraph 45), delivered after the hearing in the present case.

241

It is of no relevance, in that regard, contrary to Google’s contention (see paragraph 204 above), that, in the judgment of 3 October 1985, CBEM (311/84, EU:C:1985:394), the Court of Justice applied the conditions relating to essential facilities to an ‘active’ exclusionary practice like that at issue. In that judgment, the Court was asked, in a question referred for a preliminary ruling, about a ‘refusal to supply’ and therefore confined itself to stating its view on the conditions applicable to that practice as circumscribed by the question referred (judgment of 3 October 1985, CBEM, 311/84, EU:C:1985:394, paragraphs 19 and 26). It cannot be inferred from this that the test in relation to a refusal to supply and the resulting condition of indispensability apply to all exclusionary practices covered by Article 102 TFEU, including the practice of favouring to which the present case relates, which would moreover be at odds with the Court’s interpretation in the judgment of 17 February 2011, TeliaSonera Sverige (C‑52/09, EU:C:2011:83, paragraphs 55 and 56).

242

Furthermore, Google maintains that if, as the Commission indicates in recital 651 of the contested decision, the Court has already ruled out the application of the conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), in particular on the ground that it was not necessary that the undertaking concerned transfer an asset or that it enter into agreements with persons with whom it had not chosen to contract in complying with the decision at issue (see, to that effect, judgment of 23 October 2003, Van den Bergh Foods v Commission, T‑65/98, EU:T:2003:281, paragraph 161), the owner of an indispensable asset can always end the refusal to supply by removing the asset at issue, so that that criterion is not effective, particularly as, in the present case, the contested decision would require it, in essence, to transfer a valuable asset, namely the space allocated to search results. It follows, according to Google, that the Commission erred in relying on the judgment of 23 October 2003, Van den Bergh Foods v Commission (T‑65/98, EU:T:2003:281) in order to rule out the applicability of the conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569).

243

Admittedly, as is apparent from paragraphs 219 to 222 above, what is at issue in the present case, albeit only indirectly, are the conditions of Google’s supply of its general search service through access by comparison shopping services to the general results pages.

244

However, the obligation for an undertaking which is abusively exploiting a dominant position to transfer assets, enter into agreements or give access to its service under non-discriminatory conditions does not necessarily involve the application of the criteria laid down in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569). There can be no automatic link between the criteria for the legal classification of the abuse and the corrective measures enabling it to be remedied. Thus, if, in a situation such as that at issue in the case giving rise to the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569), the undertaking that owned the newspaper home-delivery scheme had not only refused to allow access to its infrastructure, but had also implemented active exclusionary practices that hindered the development of a competing home-delivery scheme or prevented the use of alternative methods of distribution, the criteria for identifying the abuse would have been different. In that situation, it would potentially have been possible for the undertaking penalised to end the abuse by allowing access to its own home-delivery scheme on reasonable and non-discriminatory terms. That would not, however, have meant that the abuse identified would have been only a refusal of access to its home-delivery scheme.

245

In other words, it is not because one of the ways of ending the abusive conduct is to allow competitors to appear in the boxes displayed at the top of the Google results page that the abusive practices must be limited to the display of those boxes and the conditions for identifying the abuse must be defined having regard to that aspect alone. In the present case, as is apparent in particular from recital 344 of the contested decision, the practices at issue also include the relegation of competing comparison shopping services in Google’s general results pages by means of adjustment algorithms. That relegation, in conjunction with Google’s promotion of its own results, is a constituent element of those practices and moreover, according to the Commission, plays a major role in the exclusionary effect identified and is not directly linked to access to Google’s boxes on its general results page.

246

In addition, the applicability of the criteria laid down in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569) cannot depend on the measures ordered by the Commission to bring an end to an infringement. An infringement is necessarily established before the measures to end it are determined. Accordingly, the existence of the infringement and the applicability of the conditions set out in the judgment of 26 November 1998, Bronner (C‑7/97, EU:C:1998:569) cannot depend on the measures that must subsequently be taken by the undertaking in order to bring the infringement to an end.

247

It follows that the criterion relating to the need to transfer assets or enter into agreements in order to bring an infringement to an end is not effective in the context of active infringements which, as in the present case, can be distinguished from a simple refusal to supply.

248

Lastly, Google’s assertion that the Bundesgerichtshof (Federal Court of Justice) has consistently rejected complaints that a dominant company commits an abuse by treating itself more favourably than its competitors, on the basis that companies are not required to subsidise their competitors, is unfounded. As the Federal Republic of Germany indicates in its statement in intervention, that reference to a dominant undertaking ‘subsidis[ing]’ competitors was used only in a very specific context, that of intra-group financial flows characterised by the very favourable purchase pricing set by a parent company vis-à-vis a subsidiary. In any event, even if the concept of favouring had not been recognised by the German courts, that would not preclude its relevance as a basis for the Commission’s finding of an infringement of Article 102 TFEU. The Courts of the European Union cannot be bound by the case-law of national courts, be they supreme or constitutional courts (see, to that effect, judgment of 10 April 2014, Acino v Commission, C‑269/13 P, EU:C:2014:255, paragraph 114), although there is nothing to prevent the Courts of the European Union from being guided by that case-law and taking it into account in their analysis. On the contrary, it is for the national courts and authorities to apply Article 102 TFEU uniformly and in accordance with the case-law of the Courts of the European Union, since divergences between the courts and authorities of the Member States as to its application would be liable to place in jeopardy the unity of the EU legal order and to undermine legal certainty.

249

In those circumstances, the second part of the fifth plea in law must be rejected, as must that plea in its entirety.

250

Google maintains, in essence, that the Commission misstated the facts. First, Google claims that it introduced grouped product results to improve the quality of its general search service, not to drive traffic to its own comparison shopping service. Google thus explains that it was not pursuing any anticompetitive objective by introducing product results, contrary to what is suggested by the presentation of the facts in recital 386 of the contested decision.

251

Secondly, Google contends that Product Universals did not harm users but improved the quality and relevance of its results, contrary to what is stated in particular in recital 598 of the contested decision, according to which Google did not always show users the most relevant results. In short, the contested decision ignored the evidence of Google’s pro-competitive rationale for developing grouped product results, the technical solutions that improved the quality of its general search service and actual traffic developments. The facts showed that Google had a pro-competitive rationale for showing Product Universals, which improved the quality of the general search service for the benefit of users. Google argued that it improved its technologies so that it would be more competitive with regard to the parameters on which general search engines compete. The fact that Google focused on relevance is, it argues, corroborated by its cautious triggering of Product Universals, the documentary evidence and traffic data.

252

The Commission contends in particular that it did not dispute the pro-competitive rationale for developing Product Universals as such in the contested decision. The Commission states that it took issue with Google for having shown Product Universals in an eye-catching manner while at the same time results from competing comparison shopping services could appear only as generic results, without rich display features, and that the algorithms were prone to demoting them within those results (recitals 344 and 512 of the contested decision).

253

BEUC argues that Google’s real motivation was to protect and maximise its revenue by systematically reserving the most profitable portion of the screen for its own results, which it displayed with eye-catching graphical features, even if those results were not necessarily the most relevant for a given query. Kelkoo claims that Google engaged in anticompetitive conduct to exclude its competitors and promote its own comparison shopping service. Google thus implemented a deliberate exclusion strategy designed, on the one hand, to relegate its competitors by means of its adjustment algorithms and, on the other, to favour its own comparison shopping service through more favourable display and positioning. Lastly, Visual Meta states that the allegedly pro-competitive rationale for Google’s introduction of Product Universals is, in accordance with the case-law, irrelevant and that in any event, since the alleged improvement made by Google by means of Product Universals did not benefit all competing comparison shopping services, it could not improve the relevance of its results as a whole.

254

It must be observed that where the Commission examines the conduct of an undertaking in a dominant position, that assessment being an essential prerequisite of a finding that there is an abuse of such a position, the Commission is necessarily required to assess the business strategy pursued by that undertaking. For that purpose, it is clearly legitimate for the Commission to assess subjective matters, namely the motives underlying the business strategy in question (judgment of 19 April 2012, Tomra Systems and Others v Commission, C‑549/10 P, EU:C:2012:221, paragraph 19).

255

However, the existence of any anticompetitive intent constitutes only one of a number of facts which may be taken into account in order to determine that a dominant position has been abused (judgment of 19 April 2012, Tomra Systems and Others v Commission, C‑549/10 P, EU:C:2012:221, paragraph 20).

256

The Commission is under no obligation to establish the existence of such intent on the part of the dominant undertaking in order to render Article 102 TFEU applicable even though evidence of such an intent, while it cannot be sufficient in itself, constitutes a fact that may be taken into account in order to determine that a dominant position has been abused (see judgment of 30 January 2020, Generics (UK) and Others, C‑307/18, EU:C:2020:52, paragraph 162 and the case-law cited).

257

In addition, the existence of an intention to compete on the merits, even if it were established, could not prove the absence of abuse (judgment of 19 April 2012, Tomra Systems and Others v Commission, C‑549/10 P, EU:C:2012:221, paragraph 22).

258

In the present case, Google claims that it did not in any event wish to deviate from competition on the merits and maintains, in essence, that the Commission is distorting the facts by suggesting in the contested decision that there was such an anticompetitive intent underlying the practices at issue.

259

However, it is not apparent from the recitals of the contested decision cited by Google (recitals 386, 490 to 492 and 598 of the contested decision in particular), that the Commission took into account, at least as such, for the purposes of establishing the existence of the abuse concerned any ‘anticompetitive objective’ on Google’s part in ‘developing’ the technologies that led to the introduction of Product Universals.

260

On the contrary, it is apparent from the wording of Section 7.2.1 of the contested decision that the Commission took the view that the abusive conduct consisted of objective elements, namely the fact that Google ‘positions and displays, in its general search results pages, its own comparison shopping service more favourably compared to competing comparison shopping services’, conduct which, according to Section 7.2.3 of the contested decision, in conjunction with the application of general search adjustment algorithms to competing comparison shopping services, ‘decrease[d] traffic from Google’s general search results pages to competing comparison shopping services and [increased that traffic] to Google’s own comparison shopping service’. That finding was made when traffic was, according to Section 7.2.2 of the contested decision, ‘important’ for competing comparison shopping services and, according to Section 7.2.4 of the contested decision, the diverted traffic, which accounted for a large proportion of traffic to competing comparison shopping services, could not be effectively replaced by other sources.

261

Thus, as it repeatedly stated in its written submissions, the Commission considered that Google’s conduct consisted, in particular, in the combination of two objective practices: (i) the more favourable positioning and display of Google’s own specialised results within its general results pages than the positioning and display of results from competing comparison shopping services, and (ii) the simultaneous demotion of results from competing comparison shopping services by the application of adjustment algorithms. In order to establish the infringement, the Commission thus carefully compared, first, the way in which results from competing comparison shopping services were positioned and displayed (Section 7.2.1.1 of the contested decision) and, secondly, the way in which results from Google’s comparison shopping service were positioned and displayed (Section 7.2.1.2 of the contested decision), before going on to examine, thirdly, the particular circumstances of the practices in question, namely the importance of the traffic and the fact that it could not be effectively replaced, as well as the behaviour of internet users.

262

In so doing, the Commission confined itself to comparing the way in which Google positioned and displayed the search results of competing comparison shopping services and those of its own comparison shopping service, and described the economic context in which comparison shopping services competed. In the context of establishing the infringement, the Commission did not therefore take into account any anticompetitive strategy or objectives that might have been pursued by Google, as it expressly confirmed at the hearing in response to a written question put by the Court.

263

It is true that, as is apparent from the Court’s response to the first part of the fifth plea (see paragraph 175 above), the Commission found that the practices at issue departed from competition on the merits. However, that finding cannot be invalidated by Google’s alleged intention to compete on the merits by improving the quality of its general search service and the relevance of its specialised results since, as the case-law cited in paragraph 257 above shows, the mere intention to compete on the merits, even if it were established, cannot prove the absence of abuse.

264

It must be recalled that abuse of a dominant position that is prohibited by Article 102 TFEU is an objective concept (judgment of 13 February 1979, Hoffmann-La Roche v Commission, 85/76, EU:C:1979:36, paragraph 91). It must moreover be noted that, unlike Article 101(1) TFEU, Article 102 TFEU contains no reference to the aim of the practice (judgment of 30 September 2003, Michelin v Commission, T‑203/01, EU:T:2003:250, paragraph 237), even though it refers, albeit indirectly, to an anticompetitive objective.

265

Consequently, while the Commission was entitled to comment on Google’s business strategy in the context of the launch of Product Universals and to refer in that regard to subjective factors, such as the concern to correct the poor performance of Froogle, arguments alleging distortion of the facts concerning the reasons for Google’s introduction of Product Universals must – in so far as they concern grounds that were not used by the Commission as constituent elements of the infringement (the latter being summarised in paragraph 260 above) – be rejected as being ineffective in the context of the analysis of the infringement (see, to that effect, judgment of 12 December 2018, Servier and Others v Commission, T‑691/14, under appeal, EU:T:2018:922, paragraph 188 and the case-law cited).

266

Furthermore, in so far as Google claims that Product Universals did not harm users but improved the quality and relevance of its results, it must be noted that any efficiency gains derived from the practices at issue must be taken into account as possible objective justifications, and therefore are not capable of supporting the first part of the first plea, which alleges, in essence, that the Commission distorted the facts. Those arguments relating to improvements in the quality and relevance of search results will therefore be examined at a later stage of the analysis, in Section B.4.

267

In those circumstances, the first part of the first plea in law and the fifth plea in law in its entirety must be rejected as unfounded.

268

In support of its first plea, Google puts forward three lines of argument. As previously indicated, Google maintains, first, that the facts are misstated in the contested decision because Google introduced grouped product results to improve the quality of its service, not to drive traffic to its own comparison shopping service. Secondly, Google claims that the Commission erred in finding that treating Product Universals and generic results differently had involved favouring, when there was no discrimination. In the absence of discrimination, there could not have been an abuse. Thirdly, Google claims that the Commission infringed the legal standard for assessing objective justifications – by which ostensibly anticompetitive conduct may be justified in certain cases – with respect to the display of Product Universals. It argues that the Commission did not examine whether the evidence put forward by Google in relation to the benefits for users of Product Universals outweighed the alleged restrictive effects.

269

The Court considers that this third part of the first plea, relating to the objective justifications, should be addressed separately, after the third and fourth pleas concerning the anticompetitive effects have been examined, as outlined in paragraph 126 above.

270

It will be recalled, moreover, that the first part of the first plea, according to which Google claims to have introduced grouped product results to improve the quality of its service, not to drive traffic to its own comparison shopping service, as contended by the Commission, has been addressed in paragraphs 250 to 267 above. Accordingly, only the second part of the present plea will be examined in this section.

271

Google claims that the contested decision is wrong in law in so far as it is concluded that Google favoured Product Universals, because the Commission did not examine the requirements for establishing discrimination.

272

First of all, Google states that its mechanisms for generating product results and generic results treated different situations differently, and that this was for legitimate reasons. Google does not dispute that it applied different mechanisms to generate product results and generic results. On the one hand, for generic results, Google relied on ‘crawled’ data and on generic relevance signals derived from those data. On the other hand, for product results, Google relied on data feeds provided directly by merchants and on product-specific relevance signals. By applying different technologies to generic results and product results, Google argues that it did not treat similar situations differently. It treated different situations differently and it did so for a legitimate reason, to improve the quality of its results.

273

Secondly, Google claims that it then applied the same relevance standards to specialised and generic results in a consistent manner in order to rank those results in its general search pages. In that regard, Google argues that the statement in the contested decision that Product Universals received more favourable positioning and display than generic results is also wrong because the differences in treatment did not provide Product Universals with undeserved placement on Google’s general results pages. Google states that the contested decision overlooks the way in which Universal Search operates, and claims that Universal Search had enabled a consistent ranking system for all of Google’s results categories to be established. Accordingly, Product Universals would have to have ‘earned’ their place on a results page on the basis of the same relevance standards as those applied by Google to generic results. Google indicates in that respect that, thanks to Universal Search and its constituent elements, it directly compared the relevance of product results to generic results and that it did so on the basis of the same relevance standards. Thus, when a Product Universal was displayed in a good position on the general search page, it was because it was more relevant than generic results ranked below it, not because of favourable treatment.

274

In Google’s view, the response given to those arguments in the contested decision is incorrect. In the first place, the Commission wrongly stated, in recital 440 of the contested decision, that it was irrelevant whether Google held Product Universals to the same relevance standards as generic results. Google showed Product Universals only when they were more relevant than the generic results ranked below them and they could not have received favourable treatment. They deserved the positioning they received on the general results page. Ranking results on the basis of their relevance is the opposite of favouring.

275

In the second place, in Google’s submission, the Commission was wrong to assert in recital 441 of the contested decision that Google did not demonstrate that it applied the same relevance standards to Product Universals and generic results. That reasoning seeks to reverse the burden of proof. It is for the Commission to prove that Google did not apply consistent relevance standards when showing Product Universals. Otherwise, it could not establish that there was favouring.

276

In the third place, in any event, Google submits that it did demonstrate that it applied consistent relevance standards to Product Universals. In that regard, in recital 442 of the contested decision, the Commission had wrongly claimed that Google relied on just two reports from side-by-side rater evaluations to document this. Google maintains that it provided the Commission with a large body of evidence on the operation of its ranking framework and the relevance standards that it applied. Google claims that the Commission’s criticism of those two reports in the contested decision is misplaced. The Commission was also wrong to claim in recital 390 of the contested decision that, between 2009 and September 2010, Google adopted an internal policy to ensure that the Product Universal ‘would always be positioned at the top’ whenever a result from a competing comparison shopping service was ranked in the first three generic results. This, according to Google, concerns a proposal that was never implemented. Data on the positioning of Product Universals when a comparison shopping service appeared in the first three results between December 2009 and September 2010, the period during which the Commission claims the internal policy in question was in place, contradict the Commission’s claim.

277

The Commission disputes those arguments. Kelkoo submits in particular that the adjustment algorithms were not applied to Google’s comparison shopping service and that the display formats applicable to that service were not available to competing comparison shopping services. BEUC states that Google’s product search results were not determined solely by their relevance for the consumer, as there were commercial considerations underlying the processing of those results. That is at odds with the legitimate expectation on the part of consumers that Google would be neutral in the processing of results. BEUC claims that Google manipulates the search results by making results from competing comparison shopping services invisible.

278

In essence, Google denies that the practices at issue could lead to discrimination in so far as, first, it applied different mechanisms – generic search mechanisms involving ‘crawling’ and specialised search mechanisms involving the processing of data flows sent by affiliated retailers – to different situations, that is to say, generic results and specialised results, and, secondly, it applied the same relevance criteria to all results through its Universal Search technology and its Superroot algorithm. Thus, according to Google, there could not have been any discrimination since it showed Product Universals only if they were more relevant, in the light of those technologies, than the generic results, and therefore the Product Universals would have earned their positioning on merit.

279

As a preliminary point, in order to determine whether the Commission was right in finding discrimination, the Court must examine the difference in treatment which the Commission considered contrary to Article 102 TFEU as regards in particular the positioning and display of Product Universals.

280

It must be noted in that regard that the Commission concluded in Section 7.2.1 of the contested decision that the abusive conduct consisted in the more favourable ‘positioning’ and ‘display’ of results from Google’s own comparison shopping service (including Product Universals) in its general results pages than of results from competing comparison shopping services. As is apparent from that section, from recital 344 of the contested decision and from Article 1 of the operative part of that decision, the Commission thus claims that Google positions and displays its own comparison shopping service ‘more favourably’ in its general results pages than competing comparison shopping services.

281

In order to reach that conclusion, the Commission compared the way in which results from competing comparison shopping services were ‘positioned’ and ‘displayed’ on Google’s general results pages (Section 7.2.1.1) and the way in which results from Google’s comparison shopping service, in this case, Product Universals, were ‘positioned’ and ‘displayed’ on those pages (Section 7.2.1.2).

282

The Commission concluded from this that while results from competing comparison shopping services could be displayed only as generic results, that is to say, in the form of simple blue links without pictures or additional information on the products and prices (recital 371 of the contested decision), and were subject, unlike the results from Google’s comparison shopping service, to their ranking in Google’s general results pages being demoted to the bottom of the first page or to subsequent pages by certain adjustment algorithms (recitals 352 to 355), as had moreover been the case after the introduction of the Panda algorithm (recital 361), the specialised results from Google’s comparison shopping service and, for the period at issue, Product Universals, were prominently positioned within Google’s general search results (recitals 379 and 385), displayed in rich format with pictures and information on the products (recital 397) and could not be demoted by the adjustment algorithms (recital 512).

283

The Commission thus found that the differentiated treatment of results from competing comparison shopping services and results from Google’s comparison shopping service (Product Universals) favoured the latter service compared to competing services, in the light, in particular, of the importance of the traffic generated by Google’s general search engine for competing comparison shopping services (Section 7.2.2 of the contested decision), the behaviour of internet users, whose attention tended to be drawn to the most visible results on a general search results page, that is, those that are best positioned (Section 7.2.3.1 of the contested decision), and the fact that the traffic diverted by the practices was ‘not effectively replaceable’ by other sources (Section 7.2.4 of the contested decision).

284

In the first place, it follows that the differentiated treatment alleged by the Commission does not consist, as Google suggests, in the application of different search result selection mechanisms to process the search results of Google’s comparison shopping service and those of competing comparison shopping services, that is to say, mechanisms to select generic results for competing comparison shopping services and mechanisms to select specialised search results for its own comparison shopping service. Rather, it consists in the more favourable treatment in terms of positioning and display of its own specialised results compared to the results from competing comparison shopping services appearing in the generic results.

285

Thus, while Google claims that the differentiated treatment of its search results is based on the nature of the results produced by its general search engine, that is, whether they are specialised or generic results, that differentiated treatment is in fact based on the origin of the results, that is, whether they come from competing comparison shopping services or from its own comparison shopping service. In reality, Google favours the latter compared to the former, rather than one type of result compared to another.

286

Only Google’s specialised search results, the Product Universals, can appear in the boxes on Google’s general search page, displayed in rich format, and avoid demotion by the adjustment algorithms.

287

Conversely, even if the results from competing comparison shopping services would be particularly relevant for the internet user, they can never receive the same treatment as results from Google’s comparison shopping service, whether in terms of their positioning, since, owing to their inherent characteristics, they are prone to being demoted by the adjustment algorithms and the boxes are reserved for results from Google’s comparison shopping service, or in terms of their display, since rich characters and images are also reserved to Google’s comparison shopping service. Thus, even where, notwithstanding the effect of the demotion algorithms, the relevance of the results from competing comparison shopping services is such that they appear on the first page of Google’s general results page, they can never be shown in as visible and as eye-catching a way as the results displayed in Product Universals.

288

Far from being attributable to an objective difference between two types of online result, such a difference in treatment is the result of Google’s opting to treat results from competing comparison shopping services less favourably than those from its own comparison shopping service, by displaying and positioning them less visibly.

289

In the second place, it follows from the description of the practices identified by the Commission in paragraphs 280 to 283 above that Google’s argument that it applied the ‘same relevance standards’ to grouped product results and generic results and that, in essence, it showed Product Universals only when they were more relevant than results from competing comparison shopping services through its Universal Search technology and its Superroot algorithm, so that there could not have been any discrimination, must be rejected.

290

First, it must be noted that, as is apparent from recital 440 of the contested decision, the Commission does not object to Google applying or not applying the same relevance standards to both types of result, specialised product results and generic results; it objects to the fact that Google does not apply the same standards of display and positioning to competing comparison shopping services and to its own comparison shopping service, the former being disadvantaged in comparison to the latter.

291

As is apparent from paragraph 287 above, irrespective of their relevance, results from competing comparison shopping services can never receive the same treatment as the results from Google’s comparison shopping service, whether in terms of their positioning or their display, so that they are necessarily disadvantaged in competing with them.

292

Even if the result from a competing comparison shopping service should prove to be less relevant in the light of the relevance criteria set by Google’s algorithms than a result from Google’s comparison shopping service, its demotion in Google’s general results pages by the adjustment algorithms and its limited display with just a simple generic blue link, which may be below the box reserved for the result from Google’s comparison shopping service that is displayed with rich features, is not necessarily proportionate to the lesser degree of relevance claimed in the light of those criteria. In addition, even where a competing comparison shopping service’s results are more relevant in the light of those criteria, they can never, as is noted in paragraphs 286 and 287 above, be displayed in the same way or be treated in the same way in terms of their positioning, so that competition is distorted even before the user enters a product query.

293

It follows that Google’s arguments relating to the existence of common relevance standards must be rejected as ineffective.

294

In that regard, even if the reports referred to in recital 442 of the contested decision, and, moreover, the other experiments mentioned by Google, do demonstrate the greater relevance of Product Universals compared to generic results displaying results from competing comparison shopping services, they do not address the competition concern identified by the Commission in Section 7.2.1 of the contested decision and summarised in recital 440 thereof, according to which Google processes its own specialised results and the results of competing comparison shopping services using different mechanisms for display and positioning, which necessarily has the effect of disadvantaging the results from competing comparison shopping services in comparison to the results from Google’s own service.

295

Furthermore, even if the raters preferred Product Universals to the first generic results appearing on the first general results page, which moreover is not clear from the experiments in question since they indicate that the raters’ view of the usefulness of the specialised search results and generic product results is similar, that does not mean that they preferred the Product Universals to be composed exclusively of results from Google’s comparison shopping service. Nor does it mean that they preferred the results from competing comparison shopping services to be displayed less visibly or to be subject to demotion within Google’s general results pages.

296

Secondly, it must be stated, for the sake of completeness, that Google’s argument relating to common relevance standards is not only ineffective but also unfounded, as is apparent moreover from its own written submissions. As Google itself submits, in the third part of its first plea, it could not directly compare – according to parameters applicable to specialised searches such as price, stock availability or the seller’s reputation – the specialised results from its own comparison shopping service with the specialised results from competing comparison shopping services, as it did not know how the latter’s search algorithms operated and did not have access to the data feeds sent by retailers affiliated to its sites, in particular as regards price.

297

According to Google’s own account, it knew nothing about how competing comparison shopping services ranked and scored their results for product queries and, moreover, it obtained data on products listed in its own comparison shopping service from feeds provided directly by merchants rather than by ‘crawling’ websites, as for generic searches. Google did not have comparable information on results from competing comparison shopping services, whose websites have their own feeds and their own way of ranking products.

298

Admittedly, as the Commission notes in recital 440 of the contested decision, through Universal Search, Google applied ‘certain relevance standards’ to compare its own specialised results with its generic results reproducing competitors’ results pages. However, according to the explanations given by Google in its technologies report annexed to the application, Universal Search operates on the basis of statistical criteria based on user surveys. It is neither established nor even claimed, given the difficulties set out by Google itself in comparing different types of result (see paragraph 297 above), that that tool can, without prejudice to its quality, return results as reliable in terms of the selection of relevant results as a comparison carried out by Google on the basis of its own criteria applicable to specialised product searches, namely, in particular, price, stock level, popularity of the product or reputation of the seller.

299

Thus, as reliable as they may be, these surveys are, as Foundem stated at the hearing, nothing more than a statistical approximation of what the most relevant result might be. It is apparent in that regard from an internal Google document mentioned in its technologies report that ‘one of the biggest difficulties in human evaluation is that raters must interpret the user’s query; although humans are likely to be more accurate at interpreting the query than a machine, it is impossible to know perfectly what the user intended. … this evaluation merely represents the collective opinion of our rater population, which likely differs from users at large’.

300

Lastly, Google’s claim that Product Universals were only triggered in response to 23% of ‘product searches’ and that it displayed Product Universals at the top of general results pages in response to just 4% of product queries is not sufficient to cast doubt on there being a difference in treatment. Such a trigger rate cannot, in itself, be regarded as ruling out the existence of favouring, since that rate must be compared to the trigger rate in the comparable positioning (a comparable display having been ruled out) of hundreds of competing comparison shopping services which were, given their characteristics, inherently prone to being demoted to the bottom of the page by Google’s adjustment algorithms.

301

For all of the foregoing reasons, the second part of the first plea must be rejected.

302

In support of the second plea, Google puts forward three lines of argument. First, it claims that the Commission erred in finding that treating grouped product ads and generic results differently amounted to favouring, when there was no discrimination.

303

Secondly, the contested decision erred in finding that product ads in Shopping Units favour a Google comparison shopping service.

304

Thirdly, the Commission infringed the legal rules for assessing Google’s objective justifications for showing Shopping Units. In this part of its argument, Google casts doubt incidentally on the existence of favouring with respect to its own comparison shopping service, in so far as it states that Shopping Units already included results from competing comparison shopping services. It should be observed, moreover, that that argument was invoked by Google to dispute the existence of favouring in the context of the administrative procedure (recital 405 of the contested decision). Consequently, those elements of the third line of argument must be addressed after the second. However, the elements of that argument that relate to objective justifications will be addressed together with those relating to Product Universals, after the pleas relating to the effects of the practices at issue have been examined in Section B.3.

305

In the first place, Google, supported by CCIA, submits that the Commission wrongly compared the treatment of product ads, namely Shopping Units, and the treatment of free generic results, which are not the same, and therefore there could not be any discrimination. Google asserts in that regard that paid ads, including product ads, serve to fund its general search service. Therefore, Google necessarily shows them in a different way from free generic results, which is a natural consequence of its two-sided, ad-funded business model. Google states in that respect that it marks Shopping Units on the general results page as ‘sponsored’ to designate their paid character. The claim in the contested decision that the ‘sponsored’ label ‘is likely to be understandable only by the most knowledgeable users’ (recitals 536, 599 and 663 of the contested decision) is not based on any evidence. While the Commission lists 12 differences in the contested decision between product ads in Shopping Units and text ads (recitals 426 to 438 of the contested decision) to demonstrate that product ads are not an improved form of text ads, none of those differences shows that product ads are comparable to free generic results and should be positioned and displayed in the same way.

306

In the second place, Google claims that it shows Shopping Units because they provide better ads for a product query than text ads, not because it favours them. The Commission failed to show that Shopping Units did not deserve the space allocated to them on the general results pages. High-quality ads are more helpful to users and more effective for advertisers, and they enhance the value of the search service for both categories. Google shows Shopping Units only when its product ads provide better responses to a query than text ads. As a result, Shopping Units appear in response to only around 25% of product queries, which contradicts the Commission’s claim that Google ‘always’ positions Shopping Units at the top of the page (recital 395 of the contested decision). The Commission’s claim in the contested decision that Google did not demonstrate that it ‘holds the Shopping Unit to the same relevance standards that it applies to [text] ads’ (recital 441 of the contested decision) is at odds with the fact that Google established a mechanism in which it directly compares product ads against text ads. Product ads and text ads compete to appear on the basis of the same standards of relevance and value. Furthermore, empirical data demonstrate that product ads in Shopping Units are better for users and advertisers than text ads.

307

The Commission disputes those arguments.

308

BEUC advances the same arguments as those put forward in the second part of the first plea (see paragraph 277 above). Foundem argues that Shopping Units exacerbate the anticompetitive nature of Google’s conduct inasmuch as Google replaced relevance-based results with paid advertisements that are displayed depending on the profits they generate for Google. Visual Meta states that Google’s argument that it shows Shopping Units only when they are more relevant than text ads or generic results must be rejected because the fact that its ads, with their richer format, are more visible to consumers than mere ‘blue links’ is the very reason why Google cannot reserve them to its own services without depriving consumers of more relevant results from other comparison shopping services. It is precisely the fact that rich product ads are better for users and advertisers that obliges Google to display the results from other comparison shopping services in the same formats.

309

In essence, Google reiterates the arguments raised in the context of the second part of the first plea. It maintains that the discrimination identified by the Commission has not been established in so far as (i) it merely treats differently results which are, by nature, different, that is to say, free generic results reproducing results from competing comparison shopping services and the paid-for advertising, ‘product ads’, of its own comparison shopping service (Shopping Units), and (ii) it shows Shopping Units only when those product ads provide better responses to a query than text-based advertisements, that is to say, advertisements which display a link to the advertiser’s website and a short extract of text (without images or dynamic information).

310

It must be noted that, in finding that Google engaged in the favouring of results from Google’s own comparison shopping service, the Commission compared the positioning and display of Shopping Units with the positioning and display of generic results from competing comparison shopping services. The Commission found, as for Product Universals (see paragraphs 280 to 283 above), that those results were favoured because of their prominent positioning in Google’s general results pages (Section 7.2.1.2.1 of the contested decision), the Shopping Units being always positioned above the first Google general search results (recital 395 of the contested decision), and because of their richer display in boxes at the top of Google’s general results page (Section 7.2.1.2.2 of the contested decision). The Shopping Units cannot, moreover, be demoted by competing adjustment algorithms. The Commission also found, in recital 439 of the contested decision, that results from competing comparison shopping services could not appear in the Shopping Units and, therefore, be positioned and displayed in the same way as Google’s specialised results unless those competing comparison shopping services changed their business model by becoming merchant sites offering ads in order to be eligible to appear in the Shopping Units and therefore selling products themselves.

311

In the first place, it is appropriate to address Google’s argument that its product ads in the Shopping Units constitute advertising and are thus by their very nature different from the free results of competing comparison shopping services, which would rule out any discrimination.

312

It should be pointed out that Shopping Units display results from Google’s comparison shopping service and are in competition with competing comparison shopping services. It is, in that respect, immaterial that sellers must pay an advertising fee to place products in the Shopping Units, since, for internet users, Google’s specialised search service offers the same comparison shopping service free of charge as that of competing comparison shopping services, as is apparent from Section 5.2.2 of the contested decision. Google does not therefore show how the comparison shopping service offered to internet users by the Shopping Units is intrinsically different from that offered by other comparison shopping services. On the contrary, it appears that both are designed to compare products on the internet and that, therefore, they are substitutable from the point of view of internet users.

313

It is true that Shopping Units include a ‘sponsored’ label, which is intended to indicate to internet users that they constitute advertising. However, as is apparent from recitals 536, 599 and 663 of the contested decision, the ‘sponsored’ label is not readily understood by the majority of internet users as meaning that results from Google’s comparison shopping service and results from competing comparison shopping services are ranked according to different mechanisms and that, therefore, those competing comparison shopping services may be demoted and displayed in a less visible manner in the general results pages, not because of the lesser relevance of their results compared to those from Google’s comparison shopping service, but simply because they are not Google’s own results. That is particularly so where their reduced ranking in Google’s general results pages is the product of adjustment algorithms rather than the lesser relevance of their content in relation to the query entered by the internet user, since users are not aware of that mechanism.

314

Contrary to Google’s contention, in criticising the positioning and display of Shopping Units, the contested decision does not take issue with its two-sided business model under which free services are funded by advertising. If the funding model of an undertaking leads it, as in the present case, to take part in an abuse of a dominant position, there is nothing to preclude that funding model being caught by the prohibition under Article 102 TFEU. It is, moreover, a feature of many abuses of a dominant position that the aim is to improve an undertaking’s sources of funding.

315

Google is also wrong to claim, by extension, that the Commission is calling into question the legality of its text ads, which form the basis of its business model and account for its commercial success and with which the Commission has never taken issue. Unlike Shopping Units, text ads are not part of Google’s comparison shopping service and are not being challenged for having harmed competitors in the context of a practice of favouring.

316

In the second place, to the extent that Google disputes the existence of the favouring identified by the Commission in so far as it claims to display Shopping Units only when its product ads provide better responses to a query than the text ads referred to in paragraph 309 above, and thus claims that there is no discrimination, it must be recalled that the difference in treatment identified by the Commission exists, as is apparent from paragraph 310 above, between generic results reproducing results from competing comparison shopping services and the specialised results from Google’s comparison shopping service shown in Shopping Units. Accordingly, the relevant comparison for the purpose of determining whether there has been discrimination is not between Shopping Units and text ads, but between Shopping Units and results from competing comparison shopping services that may be included in generic results.

317

It is true that, in response to an argument advanced by Google that is summarised in recital 406 of the contested decision, the Commission indicates, in recital 440 of that decision, that it is irrelevant whether Google applied common relevance standards to Shopping Units and other product ads. The Commission adds, in recital 441, that in any event Google did not apply those common relevance standards to the various types of ad.

318

However, that assessment does not alter the fact that, as is apparent from paragraphs 310 and 316 above, the difference in treatment at issue in the contested decision does not concern product ads displayed by Google, other than those appearing in the Shopping Units; rather it concerns the generic results that reproduce results from competing comparison shopping services. As is apparent from recital 440 of the contested decision, ‘the Commission … object[s] … to the fact that Google’s own comparison shopping service is not subject to those same standards as competing comparison shopping services’.

319

As is apparent from recital 439 of the contested decision which precedes recital 440 of that decision, and from paragraph 310 above, competing comparison shopping services are not eligible for the same display criteria as Google’s comparison shopping service – even by paying – to appear in Shopping Units, unless they change their business model, as is explained in paragraph 346 et seq. below.

320

It follows that the first part of the second plea must be rejected.

321

Google claims that the Commission erred in finding that product ads in Shopping Units ‘benefit[ed]’ Google’s comparison shopping service. In fact, they neither linked to that service nor generated revenue for it. The contested decision expressly recognised this.

322

Google states in that regard that the Commission listed eight reasons in the contested decision, purporting to explain why the display of Shopping Units is a means of favouring the Google Shopping ‘website’ (recitals 414 to 421 of the contested decision). However, seven of them did not identify any benefit that the Google Shopping website derives from product ads in Shopping Units, let alone a benefit that would justify counting clicks on product ads as traffic to the Google Shopping website. The contested decision mentioned, in particular, header links and ‘view all’ links in Shopping Units which do lead to the Google Shopping website (recital 419 of the contested decision). However, that does not justify objections to product ads in Shopping Units, nor does it provide a reason for counting clicks on product ads as traffic to the Google Shopping website. The Commission also noted that clicks on product ads in Shopping Units and on the Google Shopping page can lead to the same sellers’ websites (recital 418 of the contested decision). That explains the benefit of product ads – irrespective of their source – for sellers, but not how the Google Shopping website benefits from clicks on product ads in a Shopping Unit. The other reasons given in the contested decision (recitals 414 to 417 and 420) also fail to demonstrate how the Google Shopping website benefited from clicks on product ads in Shopping Units.

323

Google submits that the only reason given in the contested decision for counting clicks on product ads in Shopping Units as traffic to the Google Shopping website is that those clicks trigger a payment to Google (recitals 421 and 630 of the contested decision). However, that assertion is incorrect, because revenue generated by product ads in Shopping Units does not accrue to the Google Shopping website. Google allocates revenue from product ads in Shopping Units to its general search service. The Commission moreover acknowledged this in recital 642 of the contested decision in which it observed that Google’s display of Shopping Units ‘serves to finance its general search service’.

324

The claim that the revenue from clicks on product ads in Shopping Units benefited Google’s comparison shopping service is therefore wrong on the facts, according to Google. The reasoning in the contested decision is also wrong in law because, in essence, it raises a claim of cross-subsidisation on the ground that Google subsidises the Google Shopping website with revenues from product ads on general results pages. However, even if revenues from Shopping Units were to accrue to the Google Shopping website, which is not the case, that would not provide a basis for finding an abuse.

325

The Commission contends that Shopping Units are part of Google’s comparison shopping service, that prominently displaying Shopping Units is one means by which Google favours it, that each click on Shopping Units benefits Google’s comparison shopping service, notwithstanding the fact that those clicks lead the internet user to sellers’ websites and not to the standalone specialised Google Shopping page, and that even if revenue generated by product ads in Shopping Units does not accrue to the Google Shopping website, Google presents Shopping Units and the standalone Google Shopping page as a single service or experience to sellers and users. For sellers and users, the allocation of Google’s revenue is irrelevant (recital 420 of the contested decision). Google attempts to link the identification of the benefits for its comparison shopping service to the way in which revenue generated by clicks on Shopping Units is allocated, without taking account of the various other benefits it derives from clicks on Shopping Units, identified in recitals 445, 447 and 450 of the contested decision. The Commission adds that recitals 414 to 420 of the contested decision contain seven reasons that support the finding that clicks on Shopping Units favour Google’s comparison shopping service.

326

In relation to those aspects, Visual Meta contends, in particular, that the internal allocation of Google’s revenue cannot allow it to evade a finding of abuse within the meaning of Article 102 TFEU. Visual Meta also agrees with the Commission’s analysis in recital 630 of the contested decision, according to which Google’s comparison shopping service benefits ‘economically’ from clicks on Shopping Unit links in the same way as it would if the user had taken the intermediate step of visiting the standalone Google Shopping website and clicking on the product of the merchant partner. It states that, as is apparent from recital 421 of the contested decision, the links in Shopping Units and in Google Shopping fulfil the same economic function. Foundem and Twenga put forward essentially the same arguments.

327

As a preliminary point, it must be noted that Google’s arguments are based on the incorrect premiss that the Commission’s complaint is that Google favours its own comparison shopping service, understood to mean the standalone website corresponding to the specialised Google Shopping page, through the more favourable display and positioning of Shopping Units.

328

However, comparison shopping services are defined in recital 191 of the contested decision as being specialised search services that (i) allow internet users to search for products and compare their prices and their characteristics across the offers of several different online sellers and merchant platforms, and (ii) provide links that lead (directly or via one or more successive intermediary pages) to the websites of such online sellers or merchant platforms. That definition is not disputed by Google.

329

Consequently, recitals 26 to 35 of the contested decision must be considered to provide sufficient grounds for concluding that Google’s comparison shopping service has taken several forms, that is to say, a specialised page, most recently called Google Shopping, grouped product results, which evolved into the Product Universal, and product ads, which evolved into the Shopping Unit.

330

In those circumstances, the specialised pages Froogle, Google Product Search and Google Shopping as well as grouped product results, notably Product Universals, and product ads, notably Shopping Units, must be considered to form part of the comparison shopping service which Google offered to internet users.

331

Contrary to what is suggested by Google, the conduct at issue in the present case is not confined to the more favourable treatment of the specialised Google Shopping page by the favourable positioning and display of Shopping Units, nor does it concern ‘cross-subsidisation’. At issue here is the more favourable treatment of Google’s comparison shopping service as a whole, which includes Shopping Units.

332

As the Commission correctly pointed out in recital 412 of the contested decision, ‘[its] case is that the positioning and display of the Shopping Unit is one means by which Google favours its comparison shopping service’.

333

As is apparent from the Commission’s detailed findings, which Google has not rebutted, Shopping Units are intrinsically linked to Google Shopping so far as concerns their database of products (recital 414 of the contested decision), their mechanism for selecting results (recital 415 of the contested decision) and the results themselves, since they lead to the same landing page on the merchant site (recital 418 of the contested decision). Moreover, Shopping Units and Google Shopping are, as is also apparent from the material produced by the Commission in support of those findings, presented to internet users and merchants as a single service and a single experience (recital 420 of the contested decision).

334

Thus, sellers do not know whether they are paying for a click on their product offer in the Shopping Unit or on the standalone Google Shopping website (recital 417 of the contested decision), while internet users are invited to go to Google Shopping by a header link and a ‘view all’ link when navigating in the Shopping Units (recital 419 of the contested decision), so that, for both sellers and internet users, Shopping Units and Google Shopping constitute one and the same comparison shopping service.

335

More specifically, it must be pointed out that all the results displayed in the various structures mentioned in paragraph 329 above were results from Google’s comparison shopping service. Contrary to Google’s submission, a comparison shopping service does not merit that description only if it is capable of achieving a level of precision that allows different offers of the same product or model to be shown, as Google’s specialised web page did. A comparison shopping service can also show offers of several products that may match the internet user’s query, as in the case of Product Universals and Shopping Units. Everything depends both on the parameters of the comparison shopping service and on the precision of the internet user’s initial search query. Google cannot impose a general definition of a comparison shopping service based on Google’s own configuration of its specialised web page, Product Universals or Shopping Units.

336

In the present case, it is appropriate to adopt the definition of a comparison shopping service given in recital 191 of the contested decision and recalled in paragraph 328 above, which, moreover, has not been called into question by Google. In that regard, Google itself indicates in the glossary of technical terms annexed to the application that an aggregator is ‘a website that lists products and product offers from different merchants, and allows users to search for and compare between them’ and makes clear that ‘the Decision’ calls those sites ‘comparison shopping services’.

337

On that basis, the specialised pages Froogle, Google Product Search and Google Shopping as well as grouped product results, notably Product Universals, and product ads, notably Shopping Units, must be considered to form part of the comparison shopping service which Google offered to internet users. In addition, in relation to Shopping Units specifically, the Commission pointed out in recitals 414 to 421 of the contested decision that the Shopping Unit was based on the same database as the specialised page, that their technical and seller relations infrastructure was very largely the same, that sellers had to accept that their offers would be displayed in both and were not informed as to which of the two the clicks for which they were billed came from, that the system of payment by sellers was the same and that the internet links in the Shopping Unit and the specialised page both led to the same web page on the seller’s site. Consequently, a click in a Shopping Unit was indeed to be regarded as a manifestation of the use of Google’s comparison shopping service from the general results page, that is to say, as traffic for that comparison shopping service from that page.

338

It must be stated that certain formulations in the contested decision, such as those in recitals 408 and 423, can, viewed in isolation and at first sight, appear ambiguous. However, those formulations do not affect the Commission’s general analysis, according to which Google’s comparison shopping service was available in different forms. In particular, recital 423 of the contested decision must be read as following on from recitals 414 to 421, which are intended to show that Shopping Units and Google Shopping are components of a whole. In that regard, it must be noted that recital 422 indicates that, in six EEA countries, during a certain period, ‘Google Shopping existed only in the form of the Shopping Unit without an associated standalone website’.

339

In those circumstances, the Commission was fully entitled to find that Shopping Units favoured Google’s comparison shopping service, irrespective moreover of whether or not they favoured the standalone Google Shopping website directly by providing it with revenue.

340

Consequently, the second part of the second plea, according to which the product ads shown in the Shopping Units do not benefit Google’s comparison shopping service, must be rejected.

341

Google submits that it already includes product ads from comparison shopping services in the Shopping Units; accordingly it cannot be accused of favouring its own comparison shopping service. It states that it organises the product ads submitted by comparison shopping services with its cataloguing and indexing systems and performs the same quality controls as those applied to ads from other advertisers.

342

According to Google, several comparison shopping services in Europe – including Idealo, Twenga, Ceneo, Check24, Heureka and Kelkoo – successfully use these opportunities, placing millions of product ads on Google. It takes issue, in that respect, with the assertion made in the contested decision in recitals 344 and 371 that ‘competing comparison shopping services can appear only as generic search results’. In reality, the Commission does not dispute the fact that comparison shopping services can participate in Shopping Units.

343

On the contrary, Google argues, the Commission pointed in its letter of facts to the way in which Microsoft’s general search engine, Bing, showed product ads and to a remedy proposal from Kelkoo as a means of ending the alleged infringement. Yet both approaches corresponded to what Google was already doing.

344

According to Google, the Commission complained that accessing the Shopping Unit required comparison shopping services to change their business model by adding purchase functionality or to act ‘as intermediaries’ (recital 439 of the contested decision). However, the Commission did not explain or substantiate that complaint in the contested decision. It did not identify the specific concerns regarding the conditions that comparison shopping services must satisfy in order to participate in Shopping Units and failed to explain why those conditions are incompatible with competition law.

345

The Commission disputes those arguments. BDZV indicates that competing comparison shopping services cannot appear in Shopping Units, as that would require them to set up a ‘Google Merchant Center’ account, which requires being a merchant, that is, according to Google’s guidelines, allowing products to be purchased directly on their website. BDZV observes that comparison shopping services direct internet users to sellers’ websites. As regards the two options for comparison shopping services to appear in Shopping Units (adding a ‘buy’ button or becoming a sellers’ intermediary), BDZV explains that they fundamentally change the business model and are unlikely to convince sellers to entrust their sales to comparison shopping services, since sellers wish, in principle, to retain control over their relationships with customers. That is why only a very small number of comparison shopping services have been able to use the Shopping Units.

346

In the contested decision, the Commission stated, in recital 439, that competing comparison shopping services were not eligible to participate in Google Shopping unless they changed their business model, either by adding a ‘buy’ button or by acting as intermediaries for placing sellers’ paid product results in the Shopping Unit.

347

It is thus apparent from recital 220(2) of the contested decision that Google indicated to the Polish comparison shopping service Ceneo that it could participate in Google Shopping, and accordingly appear in the Shopping Units, only if it emulated the characteristics of online retailers or merchant platforms (the main customers of Google Shopping), either by introducing direct purchase functionality and making ‘[its website] look like a shop’, or by ‘submitting items to Google on behalf of individual merchants’ for display in the Shopping Unit and on the condition that the landing page ‘cannot give the impression of being a comparison [shopping] site’.

348

Consequently, as is apparent from recitals 439 and 220 of the contested decision mentioned in paragraphs 346 and 347 above, competing comparison shopping services were not, as such, eligible to appear in the Shopping Units. As Google confirms in the application, they can be included only if they change their business model by adding a ‘buy’ button or if they act as intermediaries to submit products to Google on behalf of online sellers. It is apparent from the application and from the reply that Google does not dispute this.

349

Yet, as BDZV points out, these options fundamentally change the business model of a comparison shopping service. It is thus apparent from recital 240 of the contested decision that a direct purchase functionality distinguishes merchant platforms from comparison shopping services, from the perspective both of internet users and of sellers.

350

As is explained in recital 240 of the contested decision, the addition of a direct purchase functionality can lead to a service no longer being considered by internet users to constitute a comparison shopping service. It can also change the regulatory framework applicable to the services provided and, above all, change the relationship of the comparison shopping site with its customers. As is also apparent from recital 221 of the contested decision, the majority of large retailers are not in favour of a ‘buy’ button being added to the websites of comparison shopping services, because those retailers wish to ‘[retain] full control over their retail activities (including the merchandising strategy, the relationships with customers and the handling of the transactions)’. Maintaining sellers’ autonomy in their selling relationship with the buyers of their products who have looked at comparison shopping services is what makes comparison shopping services unique by comparison with platforms, such as Amazon, which perform the sales function themselves for the sellers connected to them, and which are viewed as competitors by sellers who place their products on comparison sites. That is, moreover, the reason why, as is also apparent from recital 241 of the contested decision and as BDZV points out, only a very limited number of comparison shopping services have introduced that functionality (7 of the 361 competing comparison shopping services identified by Google) and, even among that limited number, the functionality concerned was introduced for only a very limited number of sellers and product offers. According to that recital, Idealo, the largest comparison shopping service after Google Shopping in Germany, had in 2015 only managed to convince less than 5% of its online sellers to add a ‘buy’ button.

351

Furthermore, the alternative offered to competing comparison shopping services in order for them to appear in Shopping Units, namely to act as intermediaries, also requires them to change their business model in that their role then involves placing products on Google’s comparison shopping service as a seller would do, and no longer to compare products. Accordingly, in order to access Shopping Units, competing comparison shopping services would have to become customers of Google’s comparison shopping service and stop being its direct competitors.

352

That assessment is not invalidated by Google’s argument that it already applies the approach taken by Bing to displaying product ads and the remedy proposed by Kelkoo to end the alleged infringement. Google claims that, as in the case of its own product ads, namely Shopping Units, product ads displayed by Bing must link to pages where users can purchase the offer concerned, and that, as proposed by Kelkoo, it receives data from third-party comparison shopping services through feeds and then organises those feeds with the aid of its own algorithms.

353

However, first, Google does not demonstrate in its written submissions that it has applied the approach advocated by Kelkoo. Kelkoo, moreover, disputes it and has placed on the file a document showing how the results of competing comparison shopping services should be treated if they are to be given equal treatment. Kelkoo indicates in that regard that, in the statement referred to by Google in the application, according to which Kelkoo is said to have conceded that Google had already applied the corrective measures requested, far from recognising that the competition concerns raised by the Commission were resolved, Kelkoo criticised Google’s assertion that Google could not apply the same processes and methods to its own results and to competing results. While Google does indeed – as was suggested by Kelkoo in its comments on Google’s response to notification of the objections – allow retailers to send it feeds containing an inventory of their products, in order to be eligible for this, comparison shopping services must change their business model, as explained in paragraph 348 above, which does not in any event address Kelkoo’s concern.

354

Secondly, even if, as Google indicates in the application, ‘Bing’s product ads must link to pages where users can purchase the offer’, that does not address the competition concern identified. What is at issue in the present case is not Microsoft’s conduct via its Bing search engine, which, moreover, is not in a dominant position on the market for general search services, but Google’s conduct. The fact that Bing’s ads also link internet users to merchants does not preclude Google’s conduct from being anticompetitive.

355

In those circumstances, the arguments put forward by Google in the third part of the second plea, according to which competing comparison shopping services were already included in the Shopping Units and therefore there could not have been any favouring, must be rejected.

356

The third and fourth pleas both concern the effects of the practices at issue. The third plea challenges their material consequences, as presented by the Commission, for traffic from Google’s general results pages to the various comparison shopping services. The fourth plea challenges the suggestion that those practices had an anticompetitive impact on the various markets identified. These aspects are linked. As indicated in paragraphs 65 to 67 above, in essence, the Commission found in the contested decision that the practices at issue had altered that traffic, which had given rise to anticompetitive effects of various kinds on the relevant markets. In those circumstances, failure to demonstrate material consequences for that traffic would necessarily mean that the premiss for a finding of anticompetitive effects on the relevant markets is missing. Likewise, the degree of importance of any material consequences for that traffic that may be demonstrated has an impact on whether or not anticompetitive effects on the markets are established.

357

It is therefore necessary to begin by examining Google’s statements concerning the material consequences of the practices at issue for traffic from its general results pages to the comparison shopping services, including its own, and then to examine Google’s argument that those practices do not have any anticompetitive effects.

358

In the first part of the third plea, Google argues that the Commission was wrong to find, in Section 7.2.3.2 of the contested decision, that the practices at issue had ‘led to a decrease in generic search traffic’ to almost all competing comparison shopping services ‘on a lasting basis’ (recital 462). Although the Commission presented numerous graphs showing the evolution of search traffic from Google to competing comparison shopping services, it failed to establish any causal link between that evolution and the practices at issue. CCIA also claims that no such link was established. The Commission should have demonstrated that the decrease it identified was attributable to the positioning and display of Product Universals and Shopping Units. The Commission could not simply presume a causal link, as is made clear by the judgment of 6 December 2012, AstraZeneca v Commission (C‑457/10 P, EU:C:2012:770, paragraph 199).

359

According to Google, supported in that regard by CCIA, the Commission was required to conduct a counterfactual analysis and examine how Google’s search traffic would have developed had the practices at issue concerning the positioning and display of Product Universals and Shopping Units not been adopted. In the contested decision, the Commission attributed decreases in search traffic from Google’s general results pages to competing comparison shopping services to other practices, which it recognised as lawful, namely the changes made by the installation of adjustment algorithms demoting certain types of website in the ranking of generic results. Contrary to the Commission’s assertion in the defence, the counterfactual analysis should not be based on a scenario in which Google no longer uses adjustment algorithms for generic results which are liable to demote comparison shopping services, since those algorithms are not in issue, which Google reiterates in its observations on a number of statements in intervention, for example that of Kelkoo, which criticises such algorithms. Neither of the two alternatives offered to Google to comply with the contested decision, set out in the Commission’s defence, namely to discontinue Shopping Units or to include competing comparison shopping services in them, involves withdrawing those algorithms. CCIA observes in that regard that the appropriate counterfactual scenario requires only a situation without the alleged abuse; in other words, the situation in which Product Universals and Shopping Units are discontinued, but not the changes made in the ranking of generic results. In response to the argument put forward by Foundem in its statement in intervention to the effect that it would be absurd for Google to withdraw product results or product ads without also withdrawing its adjustment algorithms which are capable of demoting competing comparison shopping services in the generic results, Google states that this is what it does in many countries, including in Europe, which it argues demonstrates that its proposed counterfactual analysis is not hypothetical and that those algorithms can be explained only by concerns for the quality of its results.

360

According to Google, two sets of facts relied on by the Commission should properly result in decreases in traffic to competing comparison shopping services being attributed to the changes in generic ranking by the adjustment algorithms and not to the positioning and display of Product Universals and Shopping Units. Thus, it is apparent from recitals 464 to 474 of the contested decision that none of the competing comparison shopping services mentioned in those recitals claims that the display of Product Universals and Shopping Units caused traffic losses. On the contrary, some of them expressly rejected such a link. Similarly, the second set of facts invoked by the Commission in recitals 475 to 477 of the contested decision concerns changes in the visibility of competing comparison shopping services in the generic results ‘following the introduction or update of the Panda algorithm’. The contested decision also contains, in other recitals, the assessment that the visibility of those comparison shopping services dropped ‘after the launch of the Panda algorithm’, or similar assessments (recitals 361, 367, 513 and 514), even though Google’s ranking of competing comparison shopping services in the generic results, including the application of adjustment algorithms such as Panda, is not part of the practices considered to be abusive.

361

In that regard, it is argued that recital 661 of the contested decision states that the practices at issue consist solely in the fact that Google ‘does not apply’ its adjustment algorithms for generic results (specifically, Panda) to Product Universals and Shopping Units. That is said to be readily apparent from the definition in the contested decision of the geographic scope and duration of the alleged abusive conduct, which covers only countries or periods concerned by the use of Product Universals or Shopping Units. This is why today, according to Google, since Product Universals have been discontinued, the mere removal of Shopping Units would end the infringement identified by the Commission.

362

In Google’s submission, a proper counterfactual analysis confirmed that the practices challenged by the Commission did not, in themselves, have any impact on traffic from Google’s general results pages to competing comparison shopping services.

363

Thus, first, that traffic developed similarly in countries where Product Universals and Shopping Units had been introduced and in those where they had not. Google produces in that regard a ‘difference-in-differences’ analysis, involving a counterfactual scenario of countries where Product Universals and Shopping Units were not introduced or were introduced later. Google thus compares the situation between 2004 and 2014 in the United Kingdom and Ireland, in Germany and Austria, in France and Belgium, and in the Netherlands and Belgium, in each instance for 10 or so comparison shopping services competing with Google’s own service, active in both of the countries being compared. The comparison is illustrated in the form of graphs showing traffic for each comparison shopping service in the two countries compared. For example, the evolution of traffic from Google’s general results pages to the comparison shopping service Twenga in France, where Product Universals and Shopping Units were launched, is compared to the evolution of that traffic in Belgium, where they were not. Although traffic volumes may be different, in each country pair, traffic trends over time seem to be broadly similar. The Commission’s assessment of that analysis in the contested decision is claimed to be incorrect on two counts. In the first place, it wrongly indicated, in recital 520, that the analysis does not take account of the effect of general search adjustment algorithms, in particular Panda. In the second place, it wrongly indicated, in recital 521, that traffic was not evolving in the same way in the respective country pair prior to the launch of Product Universals and Shopping Units in one of those countries.

364

Secondly, traffic to competing comparison shopping services is said not to change when Product Universals and Shopping Units are removed. In 2011, the Commission asked Microsoft to conduct an experiment (the ‘Bing Answers Experiment’) which involved removing Product Universal-type search results on Bing, its search engine, for one group of users and comparing the situation with that of another group of users who continued to see those specialised results. The data from that experiment showed that displaying, or not displaying, Product Universal-type results had a trivial impact on traffic to comparison shopping services. Google conducted its own related experiment, the ‘ablation experiment’, with Shopping Units and achieved similar results. The difference between traffic to competing comparison shopping services generated by the group of users who did not see Shopping Units and that generated by the control group was a small percentage of the total traffic of those comparison shopping services, well below the level said to have been identified by the Commission as being competitively irrelevant in the statement of objections (paragraph 446) and in the contested decision (recitals 571 and 581), even when referring to a share of almost 20% of traffic received by comparison shopping services. Moreover, the Commission was wrong to claim, in recital 523 of the contested decision, that the ablation experiment also failed to take account of the effect of general search algorithms, in particular Panda.

365

As for the two calculations which the Commission performed by reusing data from the ablation experiment in order, according to the Commission, to correct that experiment, which are mentioned in recitals 524 to 535 of the contested decision, Google claims that these are incorrect. Concerning the first calculation, illustrated in Table 22 of the contested decision, there is no basis for assuming a scenario in which comparison shopping services always appear in the top four generic results, as the Commission did. Moreover, Google was not given the opportunity to comment on this calculation during the administrative procedure, in breach of its rights of defence. Concerning the second calculation, illustrated in Table 23 of the contested decision, based on a scenario of product-only searches, which the Commission treated in the same way as searches normally returning Shopping Units, the Commission ignored the fact that comparison shopping services also received significant generic traffic from many product queries for which Shopping Units did not appear. The Commission also failed to have regard to the fact that comparison shopping services received around 50% of their traffic from sources other than Google’s generic results, which is apparent from Table 24 in the contested decision. This traffic should be taken into account when assessing the traffic impact of the Shopping Unit. If it were found that the decrease in search traffic from Google was small compared to comparison shopping services’ total traffic, the decrease could not be competitively relevant. Yet the Commission simply stated, in recital 539 of the contested decision, that the traffic allegedly diverted accounted for a ‘large proportion of traffic’ to comparison shopping services, without ever demonstrating this to be the case.

366

Thirdly, in the contested decision, the Commission did not take account of broader industry developments or shifting user preferences, as illustrated by the growing popularity of merchant platforms, such as Amazon, which are alternative options for comparison shopping searches. As the popularity of merchant platforms increased, their ranking in Google’s generic results improved compared to comparison shopping services, regardless of whether they are active in the same market. A comparison of trends in traffic from Google’s generic results to merchant platforms, on the one hand, and to comparison shopping services, on the other, confirms this analysis. Since 2008, traffic to comparison shopping services has stagnated, whereas traffic to platforms has continued to grow. While, according to internal Google documents, Amazon has established itself as the ‘benchmark in search results, speed [and] quality’ for product searches, comparison shopping services have not improved their services, as is borne out by the assessments in the file.

367

The Commission, BEUC, Foundem, VDZ, BDZV, Visual Meta, Twenga, the EFTA Surveillance Authority, Kelkoo and the Federal Republic of Germany dispute Google’s arguments.

368

In essence, Google claims that the Commission failed to establish that the practices at issue had led to a decrease in traffic from Google’s general results pages to competing comparison shopping services. According to Google, that decrease in traffic, which it does not dispute, is solely attributable to the operation of the adjustment algorithms, in particular Panda, which the Commission did not call into question. Google claims that there is no causal link between its promotion of its own comparison shopping service, as the Commission alleges, and the effect identified by the Commission, namely exclusion of competing comparison services owing to the reduction in traffic from Google’s general results pages.

369

It must, however, be pointed out at the outset that, as the Commission notes and as is set out in paragraph 69 above, the conduct of Google that is challenged in the contested decision consists in the combination of two practices: first, having displayed its comparison shopping service on its general results pages in a prominent and eye-catching manner in dedicated boxes, without that comparison service being subject to the adjustment algorithms used for general searches, and, secondly, at the same time having displayed competing comparison shopping services on those pages only in the form of general search results (blue links) that tend to be given a low ranking as a result of the application of those adjustment algorithms. It must also be pointed out that Google’s comparison shopping service, like Google’s other services, never appears as a general search result.

370

The Commission made clear, in recitals 440 and 537 of the contested decision, that it did not object, per se, to the selection criteria chosen by Google, described as relevance criteria, but to the fact that the results of Google’s comparison shopping service and those of competing comparison shopping services are not treated in the same way in terms of positioning and display.

371

Likewise, the Commission made clear, in recital 538 of the contested decision, that it did not object, as such, to the promotion of specialised comparison shopping results that Google considered to be relevant, but to the fact that the same promotion effort was not made in respect of both its own and competing comparison shopping services.

372

In essence, what the Commission called into question is the combination of practices which, on the one hand, promoted Google’s comparison shopping service and, on the other, demoted competing comparison shopping services in Google’s general results pages. It follows that the effects of those combined practices cannot be analysed by isolating the effects of one aspect of those practices from those of the other.

373

As Google submits, in themselves and taken separately, neither aspect of the practices has given rise to any competition objections as far as the Commission is concerned. However, each aspect was implemented together with the other for the periods and the territories in respect of which the Commission considered that there had been an infringement of Article 102 TFEU, and it is their joint implementation, leading, according to the Commission, to combined effects, that was considered by the Commission to be anticompetitive.

374

Consequently, the analysis of effects must take into account both the effects of the application of adjustment algorithms for generic results, in particular Panda, and the effects of the promotion of Google’s comparison shopping service by means of Product Universals and Shopping Units. Therefore, contrary to what is claimed, in essence, by Google, the analysis of the effects of the practices at issue on competing comparison shopping services cannot be limited to the impact which the appearance of results from Google’s comparison shopping service in Product Universals and Shopping Units may have had on them, that being only one of the two aspects of those practices, but must also take into account the impact of the application of adjustment algorithms for generic results. As the Commission contends, comparing scenarios in the context of a counterfactual analysis, as Google does, in which only the element of the practices that relates to the display of Product Universals or of Shopping Units varies, leads to the effect of the adjustment algorithms being neutralised, because it remains the same in both scenarios for each of the comparisons made.

375

It follows that Google’s criticism of the fact that, in order to measure the effects of the practices at issue on competing comparison shopping services, the Commission took account of the impact of generic results adjustment algorithms on traffic to competing comparison shopping services from Google’s general results pages must be rejected, and that the studies which Google put forward, which are aimed only at measuring the specific impact of the appearance of Product Universals and Shopping Units on that traffic, like the ‘difference-in-differences’ analysis or the ablation experiment, are insufficient to reflect the impact of the practices at issue on competing comparison shopping services.

376

Thus, since the situation considered anticompetitive in the present case is a combination of practices, the only counterfactual scenario that Google could properly have put forward would have been one in which no element of those practices was implemented, as otherwise the combined effects of those combined practices would be only partially understood.

377

Furthermore, identifying a credible counterfactual scenario in order to analyse the effects on a market of what are assumed to be anticompetitive practices, that is to say, identifying the events that would have occurred in the absence of the practices that are being examined and identifying the situation that would have resulted, may, in a situation such as that of the present case, be an arbitrary or even impossible exercise if that counterfactual scenario does not really exist for a market that originally had similar characteristics to the market or markets in which those practices were implemented. In principle, in the case of existing competitive relationships, not just possible or potential competition, a credible counterfactual scenario must reflect an actual situation that is initially similar but whose development is not affected by all of the practices at issue. In comparing such a counterfactual scenario with the situation observed on the market to which those practices relate, the actual effects of those practices can normally be established, by isolating them from changes that are attributable to other reasons. In that respect, a counterfactual scenario, which compares two actual developments in such a situation, can be distinguished from an assessment of potential effects which, although it must be realistic, effectively describes a probable situation.

378

Therefore, in the context of the allocation of the burden of proof referred to in paragraphs 132 to 134 above, for the purpose of demonstrating an infringement of Article 102 TFEU, particularly as regards the effects of practices on competition, the Commission cannot be required, either spontaneously or in response to a counterfactual analysis put forward by the undertaking being challenged, systematically to establish a counterfactual scenario in the sense referred to above, contrary to Google’s contention. That would, moreover, oblige it to demonstrate that the conduct at issue had actual effects, which, as will be noted in more detail in paragraphs 441 and 442 below in the examination of the first part of Google’s fourth plea, is not required in the case of an abuse of a dominant position, where it is sufficient to establish that there are potential effects.

379

In order to challenge the Commission’s assessment of the potential effects of a practice on the market, or of its actual effects if these are determined by the Commission, the relevant undertaking can certainly put forward a counterfactual analysis. However, that analysis must then enable the effects of the impugned practice as a whole to be established, rather than the partial effects.

380

In the present case, while the Commission itself drew up Table 23 of the contested decision on the basis of data from Google’s ablation experiment in response to Google’s highlighting of that experiment, the Commission did not claim that it constituted a counterfactual scenario. As the Commission explains in recital 523 et seq. of the contested decision, that table takes into account only one of the two aspects of the practices at issue, measuring the particular impact of the appearance of Shopping Units on traffic from the general results pages to competing comparison shopping services. However, as is apparent from paragraph 378 above, Google cannot reasonably criticise the Commission for having failed to establish a counterfactual scenario.

381

It must incidentally be stated that Google’s argument, mentioned in paragraph 365 above, that it did not have an opportunity to comment on the calculation resulting in Table 22 of the contested decision, in breach of its rights of defence, is ineffective. That interim calculation, which then enabled the Commission to draw up Table 23 referred to in paragraph 380 above, was merely intended to respond in detail to the counterfactual analysis that Google had already put forward during the administrative procedure but which was inaccurate, as is apparent from paragraph 375 above, and was not part of the demonstration, which will be referred to below, of the effects of the practices at issue as a whole on traffic from Google’s general results pages to competing comparison shopping services.

382

Thus, in order to establish the actual or potential effects of practices which it examines, the Commission may in particular rely on other information obtained by observation of the actual evolution of the market or markets concerned by the practices. If a correlation is observed between those practices and the modification of competition on those markets, additional information, which may include for example the assessments of market participants, their suppliers, their customers, professional or consumer associations, may be capable of demonstrating the causal link between those practices and the evolution of the market. It may in some circumstances be for the undertaking concerned to put forward relevant information that may cast doubt on that causality.

383

In the present case, in Section 7.2.3.2 of the contested decision, which deals specifically with the impact of the practices at issue on traffic from Google’s general results pages to competing comparison shopping services, the Commission first of all referred, in recitals 464 to 474, to statements from nine groups that operate comparison shopping services in several of the countries concerned, such as eBay, Nextag, Twenga or Axel Springer, indicating that those comparison shopping services had experienced significant decreases in traffic from Google’s general results pages from various dates since mid-2007, even though there had occasionally been temporary increases. For example, it is stated in recital 464 of the contested decision that, between September 2009 and September 2010, eBay subsidiaries operating comparison shopping services had lost approximately 30% of such traffic in the United Kingdom, 40% in France and 55% in Germany before other reductions in traffic to one or other of their comparison shopping sites were observed. In essence, according to what is reported in the contested decision, those groups attribute those reductions to changes in Google’s generic results adjustment algorithms, in particular Panda, which manifested themselves in a drop in the Sistrix Visibility Index by the comparison shopping services concerned. The Sistrix Visibility Index is, as is indicated in footnote 398 to the contested decision, information published weekly by the company of the same name, which takes account both of the trigger rate of a website in general search results and its ranking in those results.

384

In that regard, in recital 476 of the contested decision the Commission shows, in nine graphs, the evolution of the Sistrix Visibility Index and that of traffic from Google’s general results pages (measured by the number of ‘clicks’ on generic links) to three competing comparison shopping services between 2010 and 2014 in the United Kingdom, between 2008 and 2014 in Germany and between 2010 and 2014 in France. A fairly close correlation can be identified between the two evolutions – except for idealo.de in Germany in 2014, a year in which the two lines diverge – and, overall over the period, a decline in both – again except in the case of idealo.de which, according to what is stated in footnote 575 to the contested decision on the basis of clarification provided by Google, is accounted for by the fact that the Panda algorithm has never been applied to idealo.de. Disregarding end-of-period values that are too close to zero, the decreases between the beginning and the end of the period, irrespective of interim variations, range from 2 to 1, or 50% (guenstiger.de and touslesprix.com), to 15 to 1, or 93% (dealtime.co.uk).

385

In recital 479 of the contested decision, the Commission points out that the ‘difference-in-differences’ analysis conducted by Google, charting in particular the evolution of traffic from Google’s general results pages to 10 competing comparison shopping services, in the United Kingdom, Germany, France and the Netherlands from 2004 to 2014, also indicates a decline in that traffic to the comparison shopping services concerned, particularly after the introduction of the Panda algorithm, but also in the long term. While the differential aspect of that analysis highlighted by Google is inappropriate, because it is based on an insufficient counterfactual scenario, as explained in paragraph 375 above, the raw data of that analysis, provided in Annex A90 to the application, do enable the evolution of that traffic to be gauged for the infringement periods identified by the Commission for each country, that is to say, from the time when Product Universals were introduced there. Decreases can be identified over the entire duration of those periods and in most cases are significant from 2011 onwards for the great majority of the 40 comparison shopping services studied in the ‘difference-in-differences’ analysis, albeit that they may have been preceded or interrupted by increases and are not evident for all comparison shopping services in Germany and the Netherlands.

386

In recital 481 of the contested decision, the Commission presents, in the form of graphs drawn up on the basis of data supplied by Google, the aggregated evolution from January 2004 to December 2016 of traffic from Google’s general results pages to the 361 competing comparison shopping services identified by Google (Graphs 27 to 36 of the contested decision) for the United Kingdom, France, Germany, Spain, the Netherlands, Italy, Denmark and Poland. In the United Kingdom, a significant decline in that traffic is apparent, despite interim rises, as from September 2010 (decreasing from more than 30 million clicks to fewer than 5 million). The same situation can be observed in France, as from September 2010 (decreasing from more than 60 million clicks to fewer than 10 million) and in Germany as from September 2010 (decreasing from more than 80 million clicks to fewer than 40 million). The same is true of Spain as from January 2011 (decreasing from more than 20 million clicks to fewer than 5 million). In the Netherlands, on the other hand, the decrease can be seen only from January 2015 (decreasing from 18 million clicks to approximately 10 million). Similarly, in Italy, from a peak in September 2010 to almost 35 million clicks, the curve is irregular and ends at a little over 20 million clicks, a level first attained in 2008. In the case of Italy and the Netherlands, the Commission accepts that the traffic in question has remained stable overall. As regards Denmark, traffic is tending instead to increase, unless the comparison shopping service PriceRunner is removed from the figures, as it was by the Commission, in which case traffic has been tending to decline since September 2010 (from more than 2 million clicks to approximately 500000). The same situation can be observed in Poland, where the comparison shopping service Ceneo is driving an upwards trend. If it is removed from the figures, as it was by the Commission, traffic has been tending to decline since May 2013 (from 18 million clicks to 8 million).

387

In recital 482 et seq. of the contested decision, the Commission explains that it also set up samples of comparison shopping services competing with Google’s own service in four countries that were illustrative of the long-term impact of Google’s treatment of those comparison shopping services on its general results pages, since the ‘triggering rate’ for Shopping Units on those pages was particularly high. Those countries are the United Kingdom, for which the Commission’s sample consisted of 12 competing comparison shopping services, Germany, with a sample of 9, the Netherlands, with a sample of 6, and France, with a sample of 8 competing comparison shopping services. As the graphs numbered 53 to 56 in the contested decision show, traffic from Google’s general results pages to the comparison shopping services included in those samples did in fact decline from 2011 to 2016 in the United Kingdom, Germany and France and, having been on the increase in the Netherlands until 2014, has declined since then. A study of those graphs shows more specifically a decrease of more than half in the United Kingdom and in France, a slight decrease since 2014 in Germany and a decrease of approximately one third since 2014 in the Netherlands.

388

In addition, although contained in Section 7.3.2 of the contested decision, which is specifically concerned with assessing the existence of anticompetitive effects of the practices at issue if the market for comparison shopping services also encompasses merchant platforms, the results of the second study (‘the Second Analysis’) advanced by the Commission for that purpose, the parameters and main results of which are set out in recital 612 et seq. and the detailed results of which are in Annex 1 to the contested decision, are significant. They show, for each of the 13 countries in which the Commission found an abuse of Google’s dominant position, a decrease in the proportion of traffic to competing comparison shopping services from Google’s general results pages compared to Google’s comparison shopping service and to merchant platforms, even though, in certain countries, that traffic to competing comparison shopping services increased in absolute terms. For example, that analysis indicates, for the Czech Republic between 2011 and 2016, that there was a decrease in the competing comparison shopping services’ share from 73 to 47% (increase in absolute terms in the annual number of clicks from 62.1 million to 179.6 million). For Austria, over the same period, the analysis indicates a decrease in the competing comparison shopping services’ share from 48 to 16% or from 39 to 15%, depending on the adjustments made (decrease in absolute terms in the annual number of clicks from 68.6 million to 60.9 million).

389

In the light of these various elements put forward by the Commission in the contested decision, it must be noted first of all that Google, and CCIA, do not put forward anything in their submissions to challenge the fact that, overall, traffic from Google’s general results pages to competing comparison shopping services decreased in the 13 countries in which the Commission identified an infringement. First of all, they question only the causal link between Google’s practices and those decreases. Various expressions, in paragraph 253 of the application and in paragraph 147 of the reply, respectively, illustrate that lack of challenge. Thus, Google maintains that ‘to the extent that the ranking of comparison shopping services in Google’s generic results and associated search traffic declined, that reflected user preferences shifting to merchant platforms’ and that ‘traffic decreases caused by the application of [certain algorithms] … occurred independently of the alleged abusive conduct’. Next, as regards the Second Analysis presented in recital 612 et seq. of the contested decision, to the extent that this concerns traffic from Google’s general results pages to competing comparison shopping services, Google only questions the usefulness of that analysis for assessing whether its conduct could have had an anticompetitive effect, on the ground that that analysis does not take account of alternative sources of visits to its general results pages. As is apparent from recital 626 of the contested decision and from paragraphs 351 and 352 of the application, Google does not call into question the evaluation of that traffic in itself.

390

Google argues, moreover, in challenging the causal link between the practices at issue and the decrease in traffic from its general results pages to competing comparison shopping services, that the Commission failed to take account of comparison shopping service traffic sources other than its generic results. However, that argument is ineffective in terms of support for challenging the causal link between what is considered to be anticompetitive conduct on Google’s part and the decrease in traffic to competing comparison shopping services from its general results pages only. The argument relating to other sources of traffic will be addressed in the examination of the fourth plea, since it is restated in support of that plea.

391

Google also claims that the Commission failed to take account of broader industry developments and shifting user preferences, in particular the growing popularity of merchant platforms, including for making comparison shopping searches. Google states in that regard, in essence, that those platforms have improved the quality of their services, unlike comparison shopping services, and that that is why they are preferred by users, which has given them better rankings than those given to comparison shopping services in generic results. However, even if that is a possible explanation, it is closely linked to the functioning of Google’s algorithms ranking generic results, which is, as is noted in paragraph 373 above, one component of the practices at issue.

392

Consequently, since Google has been unable to distinguish what, in the changes in the respective ranking of merchant platforms and of comparison shopping services, would be covered solely by the improvement in the quality of service of the merchant platforms compared to the quality of comparison shopping services, all things being equal, and what would be covered by the changes made to its algorithms, notably the introduction of the Panda algorithm, that explanation does not enable the causal link which the Commission identified between the practices at issue and the decrease in traffic from Google’s general results pages to competing comparison shopping services – even partly – to be called into question.

393

Furthermore, it must be noted that Google does not challenge in its arguments the causal link between the visibility of a website within its generic results, as expressed in concrete terms by the Sistrix Visibility Index, and the importance of traffic from those results to that website. Google does not therefore challenge the fact that its algorithms ranking generic results have an impact on that traffic. Yet that causal link relates directly to one of the components of the impugned practices, the generally poor ranking of competing comparison shopping services within generic results, and to the effects of that component, namely the decrease in traffic from Google’s general results pages to those comparison shopping services.

394

In those circumstances, in the light, on the one hand, of the decreases in overall traffic which are not disputed by Google and the information derived from the statements of the nine groups operating comparison shopping services, as well as the examples of traffic decreases linked to the evolution of the Sistrix Visibility Index of various comparison shopping services presented in the contested decision, and, on the other hand, Google’s failure to produce evidence to the contrary, the Commission did demonstrate that the impugned practices had led to a decrease in generic search traffic to almost all of the competing comparison shopping services.

395

It follows from the foregoing that the first part of the third plea must be rejected.

396

In the second part of the third plea, Google submits that the Commission was wrong to claim, in Section 7.2.3.3 of the contested decision, that the impugned practices increased traffic to its own comparison shopping service.

397

In the first place, Google claims that, since those practices did not lead to a decrease in traffic to competing comparison shopping services, any increase in traffic to its own comparison shopping service could not have been at their expense and exclusionary. Exclusionary practices should by their very nature enable the undertaking engaging in such practices to take sales that competitors would otherwise have made. Accordingly, Product Universals and Shopping Units only caused the market to expand as a whole, without any adverse consequences for competing comparison shopping services. In the reply Google adds that while accepting, as the Commission maintains, that traffic to competing comparison shopping services declined after the launch of the Panda algorithm, no change associated with that event could be identified in the way in which traffic to its own comparison shopping service developed, which showed that Panda may have favoured merchant platforms, but not Google’s own comparison shopping service.

398

In the second place, Google, supported by CCIA, submits that the Commission exaggerated the volume of traffic received by Google’s comparison shopping service. First, the Commission included in that traffic clicks on ads in Shopping Units, even though those clicks did not link to the specialised Google Shopping results page, but to third-party retail websites. Visual Meta’s argument that that mechanism incentivised the sellers concerned to join Google Shopping, thereby benefiting that comparison shopping service, does not appear in the contested decision. The only reason for the Commission to count clicks on product ads is its claim that Shopping Unit revenue benefits the Google Shopping website. However, as has already been argued in the context of the second plea, that is incorrect. Thus, Visual Meta is wrong to assert that revenue from Shopping Units goes directly to Google Shopping. Moreover, the Commission did not state that that was the case in the contested decision. In its observations on the statements in intervention of Foundem and Visual Meta, Google adds that there is an inconsistency in the contested decision inasmuch as it rejects the notion that Google is a unified entity while at the same time finding that one of its individual services – its comparison shopping service – is favoured by those clicks even though they trigger payments for Google in general. In that respect, Visual Meta departs from the contested decision by arguing that the internal allocation of revenue or Google’s structure is irrelevant. In the same vein, CCIA maintains that Product Universals and Shopping Units are not part of Google’s comparison shopping service, which the Commission is said to have acknowledged in recitals 408, 412 and 423 of the contested decision. Google states, for example in its observations on Foundem’s statement in intervention, that the ads in Shopping Units do not come from the specialised Google Shopping page. Their technologies, infrastructure and formats are different, which had been demonstrated to the Commission during the administrative procedure and which the Commission does not dispute. Google also maintains, in its observations on the statement in intervention of VDZ, that Shopping Units cannot be regarded as comparison shopping services any more than Product Universals can. Shopping Units do not enable different offers for the same product or model to be compared, as comparison shopping services should do, but instead suggest a range of products capable of matching the internet user’s query. During the administrative procedure, a number of participants in the procedure endorsed that view, which the Commission had taken into account in the wording of recitals 408, 412 and 423 of the contested decision mentioned above. Secondly, according to Google, the Commission was also wrong to take account of clicks on the Shopping menu link above the results page. The existence of that menu link is not one of the components of the practices identified as abusive, the impact of which alone should be assessed. Furthermore, in the defence, the Commission did not dispute that that menu link was not a search result. As a result of those two errors, the Commission had considerably overestimated the volume of traffic that went from Product Universals and Shopping Units to Google’s comparison shopping service. In reality, as shown by a chart drawn up on the basis of connection data during the infringement period, Google sent several times more traffic from its general results pages to competing comparison shopping services than it sent to its own comparison shopping service, and three times more traffic to merchant platforms.

399

In the third place, in Google’s submission, clicks on Product Universals and Shopping Units reflect their relevance and user preferences. The reasoning for the contested decision is not convincing because the Commission only observed, in recital 494, that clicks on Product Universals and Shopping Units were all the higher because their trigger rate was high. It ignored the fact that Google displays Shopping Units (and, in the past, Product Universals) on the basis of their relevance, in the same way as all search engines, and that users click on them because they are useful, not because they appear. The visibility of Product Universals and Shopping Units and the clicks they generate are the consequence of the enhanced quality of Google’s product results and product ads, as well as user preferences. Thus, Microsoft’s experiment with its search engine Bing, the ‘Bing Algo Experiment’, referred to in recitals 460 and 461 of the contested decision, showed that users react sensitively to the relevance of results. Switching less relevant results and the most relevant results in the highest positions on Bing’s general results pages showed that users notice the deterioration in quality resulting from the promotion of less relevant results and react immediately. Microsoft thus had to abort the experiment after one week. In addition, Google submits that images in Product Universals or Shopping Units make it easier for internet users to evaluate the relevance of the proposed result because they have a preview of the product for which they are searching. The consequence of this is that they readily click on those specialised results incorporating images when they consider them a priori to be useful for their search and vice versa. Studies tracking the eye movements of internet users (‘eye-tracking’ in English or oculométrie in French), which Google conducted, bear this out. Images are thus a quality aspect of Google’s product results, not an artificial aspect designed to generate clicks. Therefore, the reason that users clicked on Product Universals and Shopping Units for years is because of their relevance, not their positioning or display. The Commission had never shown the opposite to be true. While the rise of merchant platforms did not affect traffic to Google’s comparison shopping service in the same way as traffic to competing comparison shopping services, this was because Google, unlike those competing services, innovated in terms of its product results and ads in order to keep up with Amazon and other merchant platforms, not because of the positioning and display of Product Universals and Shopping Units as the Commission suggested in recital 517 of the contested decision.

400

The Commission, BEUC, Foundem, VDZ, BDZV, Visual Meta, Twenga, Kelkoo and the Federal Republic of Germany dispute Google’s arguments.

401

It must first of all be recalled that, in Section 7.2.3.3 of the contested decision, the Commission accounted for the favourable impact of the practices at issue on traffic to Google’s comparison shopping service as follows.

402

First of all, the Commission stated in recitals 490 and 491 of the contested decision that before those practices started, Google’s comparison shopping service had been unsuccessful and was losing traffic at a rate of close to 20% per year. In recital 492 of that decision the Commission stated that, following the launch of Product Universals in the United States in November 2007, Google’s traffic had doubled in one month. In recital 493 of that decision, the Commission pointed out that Google was itself vaunting the effectiveness of ads shown in Shopping Units for sellers. In recital 494 of that decision, the Commission illustrated, in the form of graphs, the correlation between the trigger rate of Product Universals, and subsequently that of Shopping Units, and traffic from Google’s general results pages to its comparison shopping service during the periods in which those types of specialised result had been used in the United Kingdom, Germany, the Netherlands and France. For example, for the United Kingdom, Graph 37 illustrates that correlation between January 2008 and January 2013 for Product Universals, and Graph 38 illustrates it between February 2013 and December 2014 for Shopping Units. It is apparent from those two graphs that traffic to Google’s comparison shopping service rose from approximately 5 million to approximately 30 million clicks per month with Product Universals, and then from approximately 30 million to approximately 120 million clicks per month with Shopping Units.

403

In recitals 495 and 496 of the contested decision, the Commission presented the comparative evolution of traffic from Google’s general results pages to the 361 competing comparison shopping services identified by Google and to Google’s comparison shopping service, respectively, and from commencement of the practices at issue in each country until December 2016 in the United Kingdom, Germany, France, Spain, Italy, the Netherlands, Denmark and Poland. For example, for the United Kingdom, Graph 45 illustrates, between January 2008 and December 2016, a decrease from approximately 25 million to approximately 5 million clicks per month for the competing comparison shopping services and an increase from zero to approximately 350 million clicks per month for Google’s comparison shopping service. It must be noted that traffic to competing comparison shopping services is presented as being stable in Italy, the Netherlands, Denmark and Poland, which is consistent with what is indicated in Section 7.2.3.2 concerning the impact of the practices at issue on traffic to competing comparison shopping services, but that Graphs 49 to 52, respectively, show a significant increase in those four countries in traffic from those pages to Google’s comparison shopping service.

404

In recitals 497 to 501 of the contested decision, the Commission provided similar particulars for each year from 2011 to 2016, by comparing the traffic from Google’s general results pages to samples of competing comparison shopping services and to Google’s comparison shopping service in the United Kingdom, Germany, the Netherlands and France. Those samples are the same as those mentioned in paragraph 387 above. With significantly higher traffic for the samples in those four countries in 2011, traffic to Google’s comparison shopping service was, in 2016, 14 times higher than traffic to the sample in the United Kingdom, more than twice as high in Germany, more than 2.7 times higher in the Netherlands and more than 4.7 times higher in France.

405

In addition, in Section 7.3.2 of the contested decision, which is specifically concerned with assessing the existence of anticompetitive effects of the practices at issue if the market for comparison shopping services also encompasses merchant platforms, the Second Analysis, previously referred to in paragraph 388 above, shows, for each of the 13 countries in which the Commission found that there was an abuse of Google’s dominant position, an increase in the traffic share of Google’s comparison shopping service from its general results pages compared to competing comparison shopping services and compared to merchant platforms. For example, that analysis indicates, for Belgium, between 2011 and 2016, an increase in the traffic share of Google’s comparison shopping service from 0 to 22% or from 0 to 24%, depending on the adjustments made. For Norway, over the same period, the analysis indicates an increase in the traffic share of Google’s comparison shopping service from 0 to 32% or from 0 to 33%, depending on the adjustments made.

406

With regard to Google’s arguments, it must be noted that, in view of the rejection of the first part of the third plea for annulment examined at this stage, the premiss of the first set of Google’s arguments, summarised in paragraph 397 above, according to which the impugned practices did not lead to a decrease in traffic to competing comparison shopping services, necessarily cannot be accepted, nor can the contingent arguments aimed at demonstrating that an increase in traffic to Google’s comparison shopping service could not have been at the expense of competing comparison shopping services. However, even if the Commission had not demonstrated a decrease in traffic to competing comparison shopping services, that first set of arguments would have to be rejected because it is not in any event capable of demonstrating that part of the traffic which Google’s comparison shopping service was able to gain by being more visible on its general results pages than competing comparison shopping services would not have been able to go to those competing services in the absence of the practices at issue; in other words, because those arguments are not capable of demonstrating that that increase was not at the expense of those comparison shopping services, whose traffic could have increased in the absence of the practices at issue, even though it was not decreasing. As regards the argument advanced in the reply to the effect that the development of traffic to Google’s comparison shopping service was unchanged after the launch of the Panda algorithm, this will be examined in paragraphs 414 to 418 below, together with the third set of arguments according to which that development is attributable to the relevance of Product Universals and Shopping Units, rather than to the practices at issue, that is to say, together with the arguments challenging the causal link between those practices and any increase in that traffic.

407

As regards the second set of arguments put forward by Google, summarised in paragraph 398 above, according to which the Commission exaggerated the amount of traffic received by Google’s comparison shopping service, it must be noted at the outset that it is apparent from footnotes 603, 604 and 606 to the contested decision and from the replies of Google and the Commission to the questions put by the Court that traffic from Google’s general results pages to its own comparison shopping service was assessed using data supplied and explained by Google. There is no disagreement between the parties as to the accuracy of the data per se, but as to the data required to be used to determine the amount of traffic received by Google’s comparison shopping service from its general results pages. In that regard, for the period in which the name ‘Google Product Search’ was being used for the specialised search and results page, the Commission used only clicks linking internet users to the specialised page of the same name, including, for certain assessments, clicks on a specialised menu link. However, for the period in which the name ‘Google Shopping’ was used for the same page, the Commission not only used clicks linking internet users to the specialised page of the same name, including clicks on the Shopping menu link appearing in the menu, but also clicks linking the internet user directly to a seller’s site from the Shopping Units. By contrast, in its action, using the chart that appears in paragraph 269 of the application, Google provides figures counting only clicks on Product Universals and Shopping Units that linked to the specialised page called Product Search, subsequently Google Shopping.

408

The Court considers that Google’s objections to the fact that clicks on Shopping Unit ads and, where appropriate, clicks on a menu link to the specialised Google Product Search or Google Shopping page, such as the Shopping menu link, were taken into account in the assessment of the traffic of Google’s comparison shopping service coming from its general results pages, must be rejected.

409

First, as has already been indicated in paragraphs 328 to 339 above, recitals 26 to 35 and 414 to 421 of the contested decision provide sufficient grounds to support the conclusion that Google’s comparison shopping service has taken several forms, that is to say, the specialised page, most recently called Google Shopping, grouped product results, which evolved into the Product Universal, and product ads, which evolved into the Shopping Unit.

410

Secondly, as regards the criticism concerning clicks on the Shopping menu link appearing in the menu above the results page, it is true that those clicks precede use of the comparison shopping service. However, they are necessarily indicative of its use, since they signify that the internet user wishes to see the specialised page of that comparison service. In the application, Google states, moreover, in paragraph 57 that around 60% of internet users access that specialised page via that menu link.

411

In addition, as the Commission explained in reply to a question put by the Court and without being contradicted by Google, the Commission has never taken clicks on that page into account simultaneously, which could have led to double counting for a single search.

412

Furthermore, the fact that the existence of the Shopping menu link on the general results page was not challenged by the Commission as being part of the anticompetitive practices does not, contrary to Google’s contention, preclude its existence being taken into account for the purpose of assessing the evolution of traffic from its general results page to its comparison shopping service. It is true that the whole of that evolution is undoubtedly not attributable solely to the conduct for which Google is criticised, and the same is true as regards the whole evolution of traffic from Google’s general results pages to competing comparison shopping services. However, in both cases, there is a correlation between that conduct and the general trend in those evolutions, and there are numerous factors that establish a causal link in that respect, as is recalled in paragraph 383 above with regard to traffic to competing comparison shopping services and in paragraph 402 above with regard to traffic to Google’s comparison shopping service.

413

In any event, as the Commission noted in its written submissions, Google has not indicated what the evolution of traffic from its general results pages to its comparison shopping service would have been if clicks on the Shopping menu link had not been counted, but only clicks in Shopping Units, although it supplied data to the Commission distinguishing those various clicks. In those circumstances, Google has not demonstrated that the Commission incorrectly reflected the evolution of that traffic that was attributable to the practices which it considered anticompetitive. Therefore, similar criticism levelled by Google at the Second Analysis presented in recital 612 et seq. of the contested decision must, in so far as that analysis concerns traffic from Google’s general results pages to its own comparison shopping service, also be rejected.

414

As regards the third set of Google’s arguments summarised in paragraph 399 above, according to which clicks on Product Universals and Shopping Units reflected their relevance and user preferences rather than the impact of anticompetitive practices, it can hardly be disputed that internet users clicked on those specialised results and those product ads primarily because they considered them likely to be useful for their product searches, notably because they contained or took account of information that was of interest for a particular specialised search.

415

This is borne out by recitals 372 to 377 of the contested decision, mentioned by the Commission in the defence, and is essentially illustrated by recital 372 when the Commission states that ‘adding images, price and merchant information to product search results increases click-through rates [on the link displayed]’. It must, in that regard, be pointed out that the Commission does not focus, in the contested decision itself, on the intrinsic performance of Google’s comparison shopping service, as is apparent from recitals 537 and 538 of the contested decision, although it found that Google did not always display the most relevant results from comparison shopping services in a prominent place on its general results page.

416

As has been indicated in paragraphs 69, 369 and 376 above, what the Commission called into question is the difference in treatment, on Google’s general results pages, between Google’s comparison shopping service and competing comparison shopping services, which enabled results from Google’s service to be highly visible, whereas results from competing comparison shopping services could appear only through generic results and were moreover often poorly placed.

417

Nor can it be disputed that if internet users clicked on Google’s specialised product results, they did so not least because those results were promoted upfront on Google’s general results pages, which represents one of the components of Google’s combined practices. Without that visibility, those results would not have been clicked on as frequently, as is clearly shown, on the one hand, by the examples of correlations between the trigger rate of Product Universals, and subsequently that of Shopping Units, and traffic in the form of clicks by internet users from Google’s general results pages to its comparison shopping service mentioned in recital 494 of the contested decision, and, on the other hand, by the matters referred to in recital 389 of the contested decision, relating to views expressed by Google according to which the positioning of Product Universals from the top to the bottom of the first page of general results significantly influenced the number of clicks on their specialised results.

418

The third set of arguments advanced by Google based on the quality of its comparison shopping service, challenging the causal link between the practices at issue and the increase in traffic from its general results pages to its comparison shopping service, must therefore also be rejected since that link has been established, even though that traffic was also able to evolve in line with changes made to that comparison shopping service.

419

The challenge to that causal link also encompasses the argument, put forward by Google in the reply and mentioned in paragraph 397 above, that the development of traffic to its comparison shopping service was unchanged after the launch of the Panda algorithm. However, while that is indeed the case, the argument in that respect also comes up against the fact that the practices at issue are a combination of practices that concern not only the generic results adjustment algorithms, including Panda, but also the way in which the specialised product results are presented.

420

In view of the foregoing, the second part of Google’s third plea must be rejected. As indicated in paragraphs 356 and 357 above, it is necessary, therefore, to examine Google’s fourth plea, according to which the practices complained of have not had an anticompetitive impact on the various markets identified, while proceeding on the basis that the material consequences of those practices for traffic from Google’s general results pages to the various comparison shopping services, including its own, are as set out in the contested decision.

421

By its fourth plea, Google submits that the Commission failed to demonstrate that the practices at issue could have had anticompetitive effects leading, in turn, both to higher prices for sellers and consumers and to less innovation. In the contested decision, in particular, the role of Google’s strongest competitors in comparison shopping, namely merchant platforms, such as Amazon, was not taken into account and no explanation was given as to the alleged effects on prices and innovation.

422

More specifically, in the first part of its fourth plea, Google claims that the contested decision is based on pure speculation about potential effects and does not examine the actual situation and development of the markets. CCIA levels the same criticism, particularly concerning the higher prices and less innovation mentioned by the Commission. Google states that, in recital 589 of the contested decision, it is thus claimed that the conduct at issue is capable of having, or is likely to have, anticompetitive effects and, in recital 593 of the contested decision, that it has the potential to foreclose competing comparison shopping services, which may lead to higher prices and less innovation as referred to above. According to Google, there is no evidence that those eventualities came to pass.

423

In Google’s submission, the contested decision is not based on proof that the conduct at issue is, by nature, anticompetitive. Referring to the judgment of 11 September 2014, CB v Commission (C‑67/13 P, EU:C:2014:2204, paragraph 58), Google maintains that the Commission was required, for that first reason, to prove the actual anticompetitive effects of that conduct. Moreover, a second reason had arisen from the judgment of 6 September 2017, Intel v Commission (C‑413/14 P, EU:C:2017:632, paragraph 139), in which the Court of Justice held that even in cases involving conduct by a dominant undertaking that is in principle abusive, the Commission could not rely solely on evidence relating to the share of the market concerned by that conduct to find that it was indeed abusive, but had to take all the circumstances into account. In the case giving rise to that judgment, the Advocate General stated that a fully-fledged analysis had to be performed (Opinion of Advocate General Wahl in IntelCorporation v Commission, C‑413/14 P, EU:C:2016:788, point 120). That was the approach taken by the Commission in the case giving rise to its decision of 24 March 2004 relating to a proceeding pursuant to Article [102 TFEU] and Article 54 of the EEA Agreement against Microsoft Corporation (Case COMP/C‑3/37.792 – Microsoft) (OJ 2007 L 32, p. 23), which was approved by the General Court in the judgment of 17 September 2007, Microsoft v Commission (T‑201/04, EU:T:2007:289).

424

Nor, it is claimed, has it been established that Google held a dominant position on the national markets for comparison shopping services, which would have meant that competition on those markets was weakened. That, in Google’s submission, is a third reason why actual exclusionary effects on those markets should have been identified.

425

According to Google, the conduct complained of consisted in improving the service provided to internet users from its general search page by displaying specialised product search results and product ads on that page, on the basis of competition on the merits. For that fourth reason, it was necessary to identify specific exclusionary effects. The Commission and the Court had done so in similar situations. Google refers to paragraph 114 of the Commission’s decision of 21 December 1988 relating to a proceeding under Articles [101] and [102 TFEU] (IV/30.979 and 31.394, Decca Navigator System) (OJ 1989 L 43, p. 27), and to the judgments of 6 September 2017, Intel v Commission (C‑413/14 P, EU:C:2017:632, paragraph 140), and of 17 September 2007, Microsoft v Commission (T‑201/04, EU:T:2007:289, paragraphs 868, 869 and 1010). In the present case, Google’s arguments on improving the services provided to consumers were well documented and the Commission should therefore have demonstrated that the anticompetitive effects outweighed the interest in such improvements.

426

Lastly, since the conduct complained of spanned many years, its anticompetitive effects ought to have been visible if it had genuinely been harmful to competition. The duration of the conduct should therefore also, for a fifth reason, have prompted the Commission to ascertain whether that was actually the case. Google and CCIA state that, in the judgment of 12 December 2018,Servier and Others v Commission (T‑691/14, under appeal, EU:T:2018:922, paragraphs 1122 to 1128), the Court held that where the impugned conduct had already been implemented, the Commission could not – except in the case of restrictions of competition by object – merely demonstrate potential anticompetitive effects; it had to demonstrate actual anticompetitive effects, otherwise the distinction between restrictions of competition by object and restrictions of competition by effect would be illusory. Although the findings of the Court were made in the context of an anticompetitive cartel, it would nevertheless be logical to apply them to alleged cases of abuse of a dominant position. In the present case, the practices complained of did not have an anticompetitive object and the Commission should therefore have followed the Court’s approach. Proof of the existence of actual effects would in any event have enabled the Commission to substantiate the likelihood of potential effects, as the Commission itself stated in paragraph 20 of its Guidance on enforcement priorities in applying Article [102 TFEU] to abusive exclusionary conduct by dominant undertakings (OJ 2009 C 45, p. 7).

427

However, according to Google, the Commission did not demonstrate any tangible effects. Section 7.2.3 of the contested decision, to which the Commission refers to explain that it took account of specific aspects of the market, examines only the development of traffic from Google’s general results pages to competing comparison shopping services, but not the overall number of visits to them. In fact the material in the file shows that Google is not able to drive up prices or restrain innovation and that competition in the markets for comparison shopping services is robust and internet users have a wide range of choice in that respect, as the Competition and Markets Authority (United Kingdom) found in a study published in April 2017, entitled ‘Online search: Consumer and firm behaviour’. So far as prices are concerned, Google states that it demonstrated that they had fallen for sellers wishing to appear in Shopping Units.

428

Furthermore, in the light of BEUC’s arguments that Google harmed consumers by reducing their ability to access competing comparison shopping services and a wider range of sellers, Google recalls, in essence, that its relevance criteria for results shown to internet users in the generic results, Product Universals or Shopping Units are objective, particularly because Universal Search is used. The Commission did not object in the contested decision to the adjustment algorithms for generic results or to those relevance criteria, and it was only the absence of competing comparison shopping services in Product Universals and Shopping Units that it identified as problematic. Consequently, BEUC puts forward a theory that was not taken up by the Commission in the contested decision. Google also makes clear that it sent billions of free clicks of traffic to competing comparison shopping services in the 10 years preceding the adoption of the contested decision and that some of them, such as Which? in the United Kingdom, a BEUC member, saw traffic from Google’s general results pages increase significantly, as did merchant platforms. The Commission did not argue in the contested decision that Google was limiting the ability of consumers to access competing price comparison services. Google puts forward studies, including those on which BEUC relied, to show that comparison shopping services are widely used by internet users. It denies being the main entry point for online searches and states that the file relating to the procedure before the Commission bears that out. One of the studies mentioned above shows that, in the United Kingdom, Germany and France, Google’s search engine is not used as the starting point for the majority of online product searches or even used at some point in most product search journeys. In addition, contrary to BEUC’s claims, small sellers appear in ads on Google’s general results pages.

429

The Commission and, in its support, BEUC, Foundem, VDZ, BDZV, Visual Meta, Twenga, Kelkoo and the Federal Republic of Germany dispute Google’s arguments.

430

In particular, Twenga and Kelkoo indicate that the decline in traffic from Google’s general results pages to comparison shopping services competing with Google’s service was accompanied by a deterioration in the quality of their own traffic, namely a decrease in the rate at which visits to sellers’ websites from those comparison shopping services were converted into purchases. In addition, traffic from Google’s comparison shopping service to sellers increased. Twenga and Kelkoo were therefore less attractive to sellers, who, moreover, would have no interest in having their product offers appear on several websites, as their own sales websites would otherwise also be demoted in the generic results by the Panda algorithm, which downgrades sites with similar content. Twenga provides some examples of sellers who decided to dispense with its services, either due to the decline in the quality of traffic from Twenga or because, having chosen to supply Google’s comparison shopping service, they did not wish to continue to appear in the results of another comparison service. Kelkoo adds that the decrease in traffic to its site from Google’s general results pages itself led to a decrease in direct traffic to its site, which, like traffic from generic results, is ‘good quality’ traffic generating good conversion rates. Direct traffic stems from an initial visit prompted by a discovery in the generic results.

431

More broadly, BEUC states that, by limiting the visibility of competing comparison shopping services on its general results pages and favouring its own comparison shopping service and its advertisements, which are used by the largest sellers, Google not only reduced competition in the market for specialised comparison shopping search services but also restricted consumers’ ability to access a wider range of sellers and curtailed the possibility for sellers to compete with each other. BEUC states that, in the judgment of 27 March 2012, Post Danmark (C‑209/10, EU:C:2012:172, paragraph 20), the Court of Justice recalled that Article 102 TFEU covered not only practices that directly caused harm to consumers, but also practices that caused consumers harm through their impact on competition.

432

Articles 101 and 102 TFEU have the same objective, namely the preservation of undistorted competition within the internal market, as is, moreover, indicated in Protocol No 27 on the internal market and competition, annexed to the Treaty of Lisbon (OJ 2010 C 83, p. 309). In that regard, the anticompetitive practices mentioned by way of examples in those two articles are similar, although Article 101 TFEU covers agreements between undertakings and Article 102 TFEU, the unilateral practices of undertakings in a dominant position.

433

The objective of undistorted competition implies that competition takes place on a fair basis that is not adversely affected either by agreements between undertakings that restrict or eliminate competition, or by the unilateral conduct of dominant undertakings that abuse their power on the market in order, also, to restrict or eliminate competition.

434

Fair competition, including on the part of an undertaking that is, or is in the process of becoming, dominant can admittedly lead to the market-driven departure of competitors (see, to that effect, judgment of 6 September 2017, Intel v Commission, C‑413/14 P, EU:C:2017:632, paragraph 133 and the case-law cited). However, anticompetitive behaviour, in particular behaviour which is unilateral and reflects a dominant undertaking’s abusive conduct and which can also result in such a departure, is prohibited.

435

However, unlike Article 101 TFEU, Article 102 TFEU does not distinguish forms of conduct that have as their object the prevention, restriction or distortion of competition from those which do not have that object but nevertheless have that effect.

436

In the context of the application of Article 101 TFEU, when faced with certain kinds of collective conduct on the part of undertakings, a competition authority which demonstrates that the conduct in question has an anticompetitive object is not required to demonstrate its anticompetitive effects in order to characterise it as unlawful. Thus, certain collusive practices, such as price fixing within cartels, are considered so harmful, and consequently anticompetitive by nature, that it is not necessary to demonstrate their actual effects on the relevant markets (see, to that effect, judgment of 11 September 2014, CB v Commission, C‑67/13 P, EU:C:2014:2204, paragraphs 49 to 51).

437

By contrast, Article 102 TFEU provides only that any abuse of a dominant position within the internal market or in a substantial part of it is to be prohibited as incompatible with the internal market. It is apparent from the case-law of the Court of Justice that the abuse of a dominant position prohibited by that provision is an objective concept that refers in particular to the conduct of a dominant undertaking which, on a market where the degree of competition is already weakened precisely because of the presence of the undertaking concerned, through recourse to methods different from those governing normal competition in products or services on the basis of the transactions of commercial operators, has the effect of hindering the maintenance of the degree of competition still existing or the growth of that competition (see judgment of 14 October 2010, Deutsche Telekom v Commission, C‑280/08 P, EU:C:2010:603, paragraph 174 and the case-law cited). The same considerations apply where the conduct in question has restrictive effects on competition in markets related to that in which the dominant position is held.

438

As regards exclusionary practices, it has been inferred from this that a practice cannot be categorised as abuse of a dominant position unless it is demonstrated that there is an anticompetitive effect, or at the very least a potential anticompetitive effect, although, in the absence of any effect on the competitive situation of competitors, an exclusionary practice cannot be classified as abusive vis-à-vis those competitors (see, to that effect, judgments of 14 October 2010, Deutsche Telekom v Commission, C‑280/08 P, EU:C:2010:603, paragraphs 250 to 254; of 17 February 2011, TeliaSonera Sverige, C‑52/09, EU:C:2011:83, paragraphs 61 to 66; and of 19 April 2012, Tomra Systems and Others v Commission, C‑549/10 P, EU:C:2012:221, paragraph 68).

439

In that context, even when conduct of dominant undertakings is in issue that is in principle anticompetitive, such as conduct designed to secure an exclusive or highly preferential purchasing relationship with customers, possibly by means of loyalty rebates (see, to that effect, judgment of 13 February 1979, Hoffmann-La Roche v Commission, 85/76, EU:C:1979:36, paragraph 89), if the dominant undertaking concerned disputes, with documented evidence, that its conduct was capable of restricting competition, the competition authority handling the case must analyse all the relevant circumstances in order to decide what the position is (see, to that effect, judgments of 17 February 2011, TeliaSonera Sverige, C‑52/09, EU:C:2011:83, paragraph 68; of 6 October 2015, Post Danmark, C‑23/14, EU:C:2015:651, paragraph 68; and of 6 September 2017, Intel v Commission, C‑413/14 P, EU:C:2017:632, paragraphs 138 and 139).

440

In that regard, where the undertaking concerned maintains that its conduct was not capable of having – even potential – anticompetitive effects, and its case is supported by information about the actual development of the market, the competition authority is required to examine whether that information is such as to have an impact on its assessment of the existence of anticompetitive effects. In the case of practices that have actually been implemented and which are, as in this instance, complex, such information is capable of constituting relevant circumstances that may or may not corroborate the existence of an infringement of Article 102 TFEU.

441

It follows from the above that, in order to find that Google had abused its dominant position, the Commission had to demonstrate the – at least potential – effects attributable to the impugned conduct of restricting or eliminating competition on the relevant markets, taking into account all the relevant circumstances, particularly in the light of the arguments advanced by Google to contest the notion that its conduct had been capable of restricting competition.

442

However, contrary to what is claimed by Google or CCIA, the Commission was not required to identify actual exclusionary effects on the grounds that Google was allegedly not dominant on the national markets for comparison shopping services, that its conduct was part of improvements in its services for the benefit of consumers and online sellers and that that conduct had lasted for many years. Such a requirement of the Commission would be contrary to the principle, confirmed by the Courts of the European Union, that the categorisation of a practice as abuse within the meaning of Article 102 TFEU cannot be altered because the practice at issue has ultimately not achieved the desired result (see, to that effect, judgment of 17 February 2011, TeliaSonera Sverige, C‑52/09, EU:C:2011:83, paragraphs 64 and 65, and, more particularly as regards the duration of that conduct, judgment of 29 March 2012, Telefónica and Telefónica de España v Commission, T‑336/07, EU:T:2012:172, paragraph 272).

443

Nor, a fortiori, was the Commission required to demonstrate that possible consequences of the elimination or restriction of competition actually manifested themselves, for example in the form of less innovation or price increases that could only be explained by the lack of competition. It is accepted in that regard that the weakening of competition is highly likely to have such consequences, as is explained in paragraphs 11 and 19 of the Guidance on the Commission’s enforcement priorities in applying [Article 102 TFEU] to abusive exclusionary conduct by dominant undertakings.

444

It must be pointed out that the arguments to the effect that the practices at issue improved the quality of the services, notably for the benefit of the consumer, which, from the point of view of the economic interest, outweighed the exclusionary effects identified, and that therefore those practices were not abusive are arguments that fall outside the scope of an examination as to whether any effects of those practices even exist. Those arguments are therefore ineffective as support for the claim that the Commission failed to demonstrate that the practices at issue had had anticompetitive effects. In the present case, they have already been examined in part in connection with the first part of the fifth plea and they will be further examined below in connection with the third parts of the first and second pleas.

445

In the present case, in the contested decision, the Commission first analysed, in Section 7.2.3, the material consequences of the practices at issue for traffic from Google’s general results pages to competing comparison shopping services and Google’s own comparison shopping service. It follows from the examination of both parts of the third plea for annulment that the Commission was fully entitled to conclude from that analysis, in respect of the various national markets for comparison shopping services concerned, first, that those practices had led to a reduction of that traffic for almost all competing comparison shopping services and, secondly, that those practices had led to an increase in traffic to Google’s comparison shopping service. Those material effects in relation to traffic from Google’s general results pages have largely been documented, as illustrated in paragraphs 383, 388 and 402 to 405 above, and it may be concluded that the Commission established actual effects that are more or less pronounced, depending on the country, but in any event significant.

446

Next, in Section 7.2.4 of the contested decision, the Commission assessed the share of traffic that competing comparison shopping services receive from Google’s general results pages as a proportion of their total traffic (Section 7.2.4.1) and went on to explain that that share could not be effectively replaced by other sources of traffic (Section 7.2.4.2).

447

As regards the first aspect mentioned in paragraph 446 above and addressed in Section 7.2.4.1 of the contested decision, the Commission supplied details, in the form of Table 24 of the contested decision, of the various sources of traffic for 13 comparison shopping services for a six-year period from 2011 (except in one case, for which the relevant period is four