Over the past few weeks, two important developments have taken place that affect news (and, more broadly, content) markets. Both these developments concern Google. First, Google published the result of an experiment it had conducted to assess whether news content is valuable to its properties. Second, Google announced that it would start providing its AI Overviews in a further nine European countries (eight EU Member States and Switzerland) in four different languages. These two developments may seem unrelated, but they are closely linked; Google’s experiment challenges the usefulness of Article 15 of the DSM Copyright Directive by attempting to demonstrate that press publications do not add value to its properties. Article 15 requires Google to request the press publishers’ authorisation in order to use news content in AI Overviews (as well as its other services).
The implications of these developments go beyond copyright rules. Google’s experiment and the roll out of AI Overviews in Europe raise interesting questions about (a) a potential abuse of dominance in news markets, (b) a potential infringement of regulatory instruments, including the Digital Markets Act (DMA), and (c) the interplay between the potential abuse of dominance and the breach of ex ante regulation.
This post examines the above aspects, based on recent case law and the DMA proceedings against Google (which are still ongoing).
1. Google’s experiment on the value of news content
Google’s experiment sought to assess (a) whether end users would go to Search if news content was not available, and (b) how any loss in end users would affect overall ad revenue. The experiment consisted in a randomised controlled trial involving 1% of end users in several EU Member States for 2 and ½ months. To measure (a) and (b), Google removed results from Search, Discover and Google News of certain press publications that were eligible under Article 15 of the DSM Copyright Directive.
Google’s main conclusion was the following: Search ad revenue did not change despite daily average users (DAUs) declining by 0.8 percent. According to Google, this is “consistent with users continuing to use Google for more commercial queries even as they used it less for news queries”. Google also found that the overall ad revenue impact across Search, Discover, Display Ads and other Google properties could not be statistically distinguished from zero, either overall or by country (see p. 1, here).
The rationale for Google’s experiment is set out as follows:
“In 2019, the European Copyright Directive (EUCD) was passed, including Article 15, the so-called ‘neighboring right’ for press publishers. The Directive […] created new rights for news publishers when extended previews of their work are used online. Even before EUCD, Google had a history of deep collaboration across the news industry globally. But to comply with its provisions, over the past few years Google rolled out a licensing program in Europe called Extended News Previews (ENP). After working with both individual publishers and collecting societies, we now have agreements covering more than 4,400 publications across 24 countries in the EU. We’re the first company to have implemented a program dedicated to EUCD compliance, from which eligible publishers of all sizes can benefit. As we’ve rolled out our ENP program, regulators and publishers across the EU and globally have made numerous requests in writing for additional data about the effect of news content in Search on people’s use of our products and on the economic value of this content to Google. In France, we said in a 2024 blogpost that we intended to engage in further analysis of the value of news content to our services:
‘To guide future negotiations, [Google] will be taking steps to further analyse the true economic value of news content in our services.’
To meet the requests of relevant stakeholders for more such data, Google ran an experiment intended to measure the value of Art. 15 EUCD press publication content in our services. ‘True economic value’ comes from the presumed increase in ad revenue that would pertain from the inclusion of news content on Google services, or equivalently, from the presumed decrease in ad revenue that would pertain from the exclusion of news content on Google services”.
Put simply, Google’s experiment sought to challenge the premise of Article 15 of the DSM Copyright Directive that “the use of press publications constitutes an important part of [platforms’] business models and a source of revenue” [recitals 54-55]. Based on the results of Google’s experiment, the assumption on which Article 15 of the DSM Copyright Directive rests is flawed.
However, there is a different way in which one can read the results of Google’s trial. If one considers how end users (a) use Search, and (b) consume news content, the main conclusion that could be drawn from the experiment is that Google is dominant and that the removal of news content from its properties -especially Search- would amount to an abuse of dominance.
Numerous studies, including the flagship annual report of the Reuters Institute, find that platforms, including search engines, are the main gateway to online news (see p. 10, here). Given that Search controls 90+% of the search market in the EU, one immediately realises the gatekeeping role that Google plays in news markets. In other words, that the removal of news content from Google’s properties did not result in a decline in its user base may not be evidence that end users go to Search for other types of content, but that Google is dominant (to the effect that it can degrade the quality of its service without users switching to an alternative).
This reading of the results of Google’s experiment is not only supported by the findings of the Reuters Institute (and other) report(s). It is further supported by the findings of competition authorities. Before we turn to recent case law, it is worth reminding ourselves of the legal definition of dominance:
“Dominance has been defined in [EU] law as a position of economic strength enjoyed by an undertaking, which enables it to prevent effective competition being maintained on a relevant market, by affording it the power to behave to an appreciable extent independently of its competitors, its customers and ultimately of consumers. This notion of independence is related to the degree of competitive constraint exerted on the undertaking in question. Dominance entails that these competitive constraints are not sufficiently effective and hence that the undertaking in question enjoys substantial market power over a period of time. This means that the undertaking’s decisions are largely insensitive to the actions and reactions of competitors, customers and, ultimately, consumers” (see paragraph 10, here)
The application of the legal definition of dominance to the facts of the Google Shopping case is particularly relevant here. To reach a conclusion on whether Google is dominant in search, the Commission’s assessment was based inter alia on the infrequency of user multi-homing, the existence of brand effects, and the lack of countervailing buyer power. The most notable findings include the following:
– Notwithstanding the technical ability of users to switch between different general search services, only a minority of users in the EEA that use Google’s general search service as their main general search service use other general search services (i.e., there is no multi-homing).
– Because of the strength of the Google brand, users trust in the relevance of search results provided by Google. Consequently, as supported by internal Google documents and two third party studies, a significant number of users are unlikely to multi-home even if Google were to degrade the quality of its general search service.
– Contrary to what Google claims, the Commission is not required to ascertain the number of users that would need to multi-home to render a possible degradation by Google in the quality of its general search service unprofitable. There is no legal requirement on the Commission to quantify how much a dominant undertaking would have to alter its product for it to no longer to be dominant.
– Users are unable to exert any meaningful countervailing buyer power vis-à-vis Google. This is because the national markets for general search services are characterised by the fact that each user only represents a tiny fraction of the volume of total search queries.
In the light of the Commission’s findings in Google Shopping, removing news content from Search may not support Google’s claim that users use Search for commercial queries but the claim that Google is dominant because it has the ability to degrade the quality of the service without fearing that consumers would switch to another service.
A hypothetical scenario involving the removal of news content from Search could further trigger a finding of abuse. As the Commission’s Guidance on Article 102 TFEU notes, dominant undertakings may impair consumer welfare “whether in the form of higher price levels than would have otherwise prevailed or in some other form such as limiting quality or reducing consumer choice.” There is case law that illustrates why such conduct could qualify as abusive. In September 2019, in anticipation of the entry into force of the French Law transposing the DSM Copyright Directive, Google announced to publishers and news agencies its new policy for displaying their content within its various services, which read as follows:
“When the French law comes into force, we will not show preview in France for a European news publication unless the publisher has taken steps to tell us that’s what they want. This applies to search results across Google services” (see paragraph 91, here).
Google also noted that it did not intend to remunerate press publishers for including their editorial content – see relevant Google FAQs:
“Q: If I use the settings to allow [Google to use my news content], will Google pay me?
A: We believe that the operation of the search engine should be based on relevance and quality, and not on commercial agreements. To do otherwise would reduce the choice and relevance of search results for our users – and undermine confidence in our services. For this reason, […] we do not remunerate links or previews in search results. When you use the new settings, you agree to the use of previews of your content without payment, both for traffic to and from Google” (paragraph 94, here).
In other words, a publisher’s agreement to the use of its content in Search was equivalent to granting Google a free licence. Otherwise, Google would not display the content in Search (to the detriment of news publishers, which would lose (a) one of their main traffic sources and (b) ad revenues).
Following Google’s unilateral decision to refrain from paying for news content (in contravention of the DSM Copyright Directive), the French competition authority (Autorité de la Concurrence or AdlC) issued interim measures requiring Google to initiate negotiations with affected publishers in order to make a financial proposal for the use of the protected content (see here). The AdlC’s decision ordering interim measures found that Google’s conduct was likely to qualify as an abuse of dominance on the grounds that: (a) Google imposed on news publishers unfair trading conditions; (b) Google engaged in anti-competitive discrimination by imposing zero remuneration on all protected publishers, and (c) Google circumvented the scope of the DSM Copyright Directive (and the national law transposing it).
The AdlC found in 2021 that Google had not complied with these interim measures. It imposed a fine of €500 million and ordered Google to comply, under penalty payment, with the initial injunctions. Google ultimately offered commitments, which the AdlC made binding in a decision of June 2022 (see here). In March 2024, the AdlC fined Google €250 million for failing to comply with some of the commitments made binding by its decision of June 2022, the aim of which was to guarantee the following principles: (a) negotiate in good faith, based on transparent, objective and non-discriminatory criteria, within three months; (b) provide press publishers with the information needed to transparently assess their remuneration for related rights; and (c) take the necessary measures to ensure that negotiations do not affect other economic relationships between Google and press agencies and publishers (see here).
The March 2024 enforcement initiative also dealt with Google’s use of third-party news content for the purposes of training Gemini (Bard at the time). The AdlC found that Bard had used content from press publishers to train its foundation model, without notifying either them or the AdlC. Google further failed to propose a technical solution for press agencies and publishers to opt out of the use of their content by Bard without affecting the display of content protected by related rights on other Google services, thereby obstructing the ability of press agencies and publishers to negotiate remuneration.
The AdlC proceedings demonstrated a gap the DSM Copyright Directive leaves open. The latter enshrines the publishers’ exclusive right to authorise (or prohibit) the reproduction or the making available to the public of their publications by information society service providers (Article 15(1)). Moreover, the DSM Copyright Directive establishes the Member States’ obligation to ensure that copyright holders are entitled to receive appropriate and proportionate remuneration (Article 18(1)). However, it does not establish an obligation for the information society service providers that fall in scope to negotiate a licensing agreement with the protected publishers. In practice, this means that a dominant platform may exert bargaining power over protected publishers by subjecting access to the platform to the granting of free licences. This is why the AdlC launched the proceedings against Google (a practice that could not be remedied by copyright rules but that could be remedied by competition rules). One of the pillars on which the AdlC relied was the case law where the EU courts held that an undertaking may be infringing Article 102 TFEU when, without formally infringing a law, it circumvents its purposes without any objective justification.
It is in the light of the Commission’s Google Shopping decision and the French proceedings that the results of Google’s experiment must be read.
Google’s experiment may have some policy implications for the revision of the DSM Copyright Directive (which is due to start in 2026). If one considers that (a) search engines are the main gateway to news content, (b) a dominant/gatekeeping search engine is in the position to remove news content without users reacting to such removal, and (c) the DSM Copyright Directive does not establish the duty of powerful platforms to negotiate, the case can be made that an obligation should be introduced in the Directive that requires companies that have been designated as gatekeepers under the DMA to negotiate in good faith with press publishers. Such a requirement would be more appropriate to achieve the objectives of the Directive, as set out in recital (54):
“A free and pluralist press is essential to ensure quality journalism and citizens’ access to information. It provides a fundamental contribution to public debate and the proper functioning of a democratic society. The wide availability of press publications online has given rise to the emergence of new online services […], for which the reuse of press publications constitutes an important part of their business models and a source of revenue. Publishers of press publications are facing problems in licensing the online use of their publications to the providers of those kinds of services, making it more difficult for them to recoup their investments. In the absence of recognition of publishers of press publications as rightholders [and in view of the bargaining power held by a handful of gatekeeping services], the licensing and enforcement of rights in press publications regarding online uses by information society service providers in the digital environment are often complex and inefficient.”
This approach would reflect similar efforts in the UK where, under the DMCC Act, it is expected that platforms that will be designated as SMS entities must pay for the (news) content they use. Finally, it is worth noting that France was not included in Google’s experiment. The Paris Commercial Court suspended it, following an action brought by the Reporters Without Borders.
2. The roll out of AI Overviews
A few days after the publication of the results of its experiment, Google announced that it started offering AI Overviews to a further nine European countries. According to Google’s description, “AI Overviews use generative AI to provide key information about a topic or question, with links to dig deeper and learn more on the web. More than a billion people around the world now use AI Overviews for help with their questions”. In practice, AI Overviews form a (text-based) content service that provides answers to a user’s query. This content service is followed by the search results users are familiar with.
In addition to the implications arising from the perspective of the DSM Copyright Directive, the provision of AI Overviews likely falls foul of competition law and the obligations established in the DMA.
-As already mentioned above, under the DSM Copyright Directive, the use of news content for the purposes of offering AI Overviews is conditional on the publishers’ authorisation.
-As regards the DMA, AI Overviews seem to breach Article 6(5), which requires gatekeepers to refrain from “treating more favourably, in ranking and related indexing and crawling, services and products offered by the gatekeeper itself than similar services or products of a third party”. AI Overviews may be regarded as a form of self-preferencing in ranking for the reasons set out below:
The display of content is a separate service
Article 6(5) rests on the distinction between the designated service (i.e., Google Search) and other services offered by the gatekeeper. Search engines are different from content sites; the former act as directories that organise content whereas the latter provide the answer to the user’s query. As the Commission found in Google Shopping, while a search engine primarily seeks to guide users to other sites, content sites offer directly the information (see Google Shopping, paragraph 164). This means that, from the users’ perspective, the provision of content and the provision of search results are distinct services. The same applies to data that Google (a) purchases for display to users, or (b) aggregates or amends to answer user queries. These parameters should guide the Commission’s interpretation of Article 6(5), which distinguishes the designated service from other services offered by the gatekeeper and third parties.
Relatedly, Article 13(4) prevents Google from engaging in any behaviour that undermines effective compliance with Article 6(5) regardless of whether that behaviour consists in the use of behavioural techniques or interface design. By answering the end users’ queries, Google uses the design of its interface to discriminate in favour of its own content service, breaching the anti-circumvention rules of the DMA.
The direct embedding of content from Google in the SERPs or the more prominent display of links to Google’s own content services constitutes self-preferencing that is caught by the DMA
Preferential treatment in ranking may take various forms, such as the direct embedding in the search results, the inclusion in a box above or on the side of the search results, and the provision of the result with or without a link to a separate page. Recital (51), which interprets Article 6(5), explicitly notes that self-preferencing can occur with services that are “partly or entirely embedded in online search engines results [and] displayed along with the results of an online search engine, which are considered or used by certain end users as a service distinct or additional to the online search engine”. The Commission’s findings in Google Shopping (see above) illustrate why end users distinguish content sites from search engines. Moreover, Recital (52) clarifies that the term “ranking” in Article 6(5) should cover all forms of relative prominence, including display and linking. In view of the above, Article 6(5) unequivocally covers the direct embedding of content from Google in the SERPs and the more prominent display of links to Google’s own content services.
Google’s AI Overviews are a separate service
Google’s AI Overviews generate content to directly answer user queries. This goes beyond the scope of a search engine and forms a separate service (to the effect that AI Overviews fall under the scope of Article 6(5) as they form the service receiving preferential treatment). It is worth noting that Article 6(5) applies to all services offered by the gatekeeper itself. This includes services that existed at the time the DMA started to apply and new services, including AI Overviews.
A few days before Google’s announcement that it started offering AI Overviews in several EU Member States, the European Commission issued a press release setting out its preliminary findings on Google’s implementation of Article 6(5). In this press release, the Commission notes that Google treats its own services, such as shopping, hotel booking, transport, or financial and sports results, more favourably in Google Search results than similar services offered by third parties. More specifically, Alphabet gives its own services more prominent treatment compared to others by displaying them at the top of Google Search results or on dedicated spaces, with enhanced visual formats and filtering mechanisms.” This statement does not refer to news content but there is no reason why news should be treated differently from e.g., sports and financial results.
In addition to breaches of the DMA, the provision of AI Overviews would also seem to be covered by Article 102 TFEU:
– Google Shopping dealt with the same issue, that is, self-preferencing in ranking.
– Another angle is the Amazon antitrust investigation (which Google settled through commitments); similar to Amazon, Google seems to be using data/content from news publishers to enhance its content offering.
– A different perspective is offered by the recent Meta Platforms judgment where the Court of Justice of the EU ruled that, in the context of examining whether an undertaking has abused its dominant position, competition authorities may need to examine whether that undertaking’s conduct complies with rules other than those relating to competition law, “such as the rules on the protection of personal data laid down by the GDPR.” The wording of the ruling suggests that competition authorities may consider the breach of rules other than data protection regulation in order to assess infringements of competition rules. Such rules (clearly) include the DMA and the DSM Copyright Directive.
Conclusion
A few years ago, the legal framework offered a limited toolkit to address the practices discussed in this blog. Competition law was perhaps the only tool that could be used to remedy the harm resulting from such practices (but antitrust proceedings can be slow, and Article 102 TFEU can only be enforced ex post, that is, after the harm materialises). The legal framework has evolved to impose obligations and prohibitions that reflect the role platforms play in the quality and variety of content end users engage with. It is now time to use these tools not only to ensure their effet utile but also (and equally -if not more- importantly) to protect a pluralist press.
Dr. Konstantina Bania