Scholars argue that Google’s use of publisher data to train AI violates U.S. and EU antitrust laws.
Every time that users see an AI Overview in response to their Google search query, a publisher somewhere loses a click. The queries that once directed users to news articles, recipes, or sports reports are now answered by Google search using content that publishers created.
In a forthcoming article, Madhavi Singh and Fiona M. Scott Morton of the Thurman Arnold Project at Yale University contend that Google’s integration of AI Overviews into its search engine constitutes illegal tying that violates the Sherman Act.
AI Overviews are AI-generated summaries that appear at the top of Google’s search results page. These overviews attempt to answer user queries, resulting in fewer users clicking through to publisher websites and, in turn, depriving publishers of advertising revenue.
Singh and Scott Morton trace the evolution of Google’s relationship with publishers from a mutually beneficial arrangement to a one-sided transaction. Initially, publishers permitted Google to crawl their websites to build its search index in exchange for Google directing traffic to publishers. The introduction of AI Overviews disrupted that arrangement. Google now uses publisher data not only for search indexing but also in its AI Overview summaries and retains users on its own page rather than directing them to publisher websites.
Singh and Scott Morton argue that Google ties inclusion in its search index to the use of publisher data for AI-related purposes. Although Google allowed publishers to opt out of AI training, publishers did not exercise this option. Most were apprehensive that opting out of Google’s AI crawlers would risk exclusion from Google search entirely.
Singh and Morton highlight that publishers were correct to be apprehensive, pointing to documents from Google’s testimony in the recent Google Search trial which revealed that the company did not honor the opt-out. Google justified its actions by arguing that AI Overview was a feature of search, that is distinct from its AI lab, for which the opt-out was honored. Singh and Scott Morton question Google’s motives in doing so, pointing to to internal documents which showed that honoring publisher opt-outs for AI Overviews would have excluded roughly half of its training data.
Singh and Scott Morton argue that Google’s conduct satisfies both elements of anticompetitive tying under the Sherman Act. First, Google is tying two distinct products, Search and AI training. Second, Google’s conduct fulfils the coercion element of anticompetitive tying, as publishers cannot refuse AI-related data uses without forfeiting search visibility. Although Google may claim that the feature improves user experience and answers queries faster, Singh and Scott Morton emphasize that harms to publishers cannot be offset by benefits to consumers
Singh and Scott Morton argue that the anticompetitive effects on publishers have become increasingly evident. Click-through rates to publishers’ websites for informational queries featuring AI Overviews fell 61 percent since mid-2024. Publishers also experienced 10 percent less traffic referred to their websites from Google Search in the sample period studied for 2025, as against the same period in 2024.
Major outlets such as CNN and Business Insider saw traffic decline by 30 percent and 40 percent, respectively, and smaller publishers suffered even greater losses. Singh and Scott Morton warn that this trajectory could culminate in “Google Zero,” where Google effectively stops sending meaningful traffic to publishers.
Singh and Scott Morton emphasize that assessing harm to publishers is key in monopsony cases, where a buyer exercises monopoly power. A federal court already found that Google illegally monopolized the search market with more than 80 percent market share. Singh and Scott Morton extend this finding, arguing that Google’s monopsony over publisher data stems from its monopoly over general search.
Beyond U.S. antitrust law, Singh and Scott Morton argue that Google’s conduct also violates EU competition law. They contend that using publisher data for additional, revenue-generating purposes without paying them additional compensation would amount to unfair pricing. Singh and Scott Morton also argue that conditioning search indexing on AI data use is an unfair trading condition under EU competition law.
Singh and Scott Morton explore remedies that courts can deploy to restore competitive market conditions. They evaluate several options: an opt-in framework, where publishers can decide how their data is used; regulatory price-setting; collective negotiations among publishers to grant them some bargaining power against Google; and establishing an industry organization to counteract Google’s monopsony. Singh and Scott Morton favor the opt-in approach as a necessary remedy and recommend strengthening publishers’ ability to negotiate as an additional measure.
Singh and Scott Morton stress the urgency of enforcement. Courts have already found that Google illegally monopolized the general search and digital advertising markets, and Singh and Scott Morton argue that further antitrust violations should not go unpunished. Publishers may face irreversible anticompetitive effects. Once independent journalism erodes, the market cannot easily rebuild. Singh and Scott Morton warn that the open web risks becoming a pool of recycled AI-generated content unless regulators act promptly.
