Scholars from Yale’s Thurman Arnold Project say, based on their recent analysis, that Google’s new AI Overviews may breach U.S. antitrust rules and EU competition standards.
What the new summaries do
When a user types a question, the search engine now displays an AI-generated paragraph immediately after the query is submitted, placing it at the very top of the results page. This brief summary is designed to answer the query directly without requiring the user to scroll further.
That placement keeps the user on the search page, reducing the need to click through to the original source. The authors note that each displayed overview replaces a potential click on a third-party site, meaning that publishers lose a chance to attract a reader who might otherwise have visited their own page.
Data from a 2025 study show click-through rates for informational queries fell 61 percent after the feature rolled out in mid-2024. Overall traffic referred from the search engine dropped about 10 percent compared with the same period a year earlier, indicating a measurable shift in user behavior.
Major outlets such as CNN and Business Insider reported traffic losses of roughly 30 percent and 40 percent, respectively. Smaller news sites saw even larger declines, according to the same analysis, which suggests that the impact is felt most sharply by organizations with less brand recognition.
Legal framing of the practice
Madhavi Singh and Fiona M. Scott Morton argue the practice ties two separate products—search results and AI training data—against the Sherman Act. They say publishers cannot refuse the data use without losing visibility in the search index.
The scholars point to testimony from a recent search-engine trial showing the company did not honor opt-out requests for AI training. While the opt-out applied to a separate AI lab, the overviews continued to draw on the same content, raising questions about compliance.
Internal documents suggest honoring opt-outs would have removed about half of the material used to train the summaries. The authors interpret that as evidence the firm prioritized the feature over contractual fairness.
They note that a federal court already found Google held more than 80 percent of the search market, establishing a monopoly. The new claim extends that finding to a monopsony over publisher data.
Implications for the news ecosystem
One way to address the imbalance would be to let content creators decide whether their material can be used for AI training. An opt-in system could restore bargaining power and align usage with compensation, giving publishers a clearer choice about participation.
Evolution of Publisher Relations and Risk of Recycled Content
The scholars describe how the relationship began as a two-way exchange: publishers allowed Google to crawl their sites to build its index, and Google sent traffic back to those sites. The rollout of AI Overviews altered that balance, turning the deal into a one-sided arrangement that limits publisher benefit.
They caution that, without timely regulation, the open web could become a repository of recycled AI-generated content. Such a shift would reduce the availability of original reporting and erode the diversity of information online.
Possible court-driven fixes
The authors outline several remedies, including regulatory price-setting, collective bargaining structures, and a new industry body to counter the firm’s leverage. They favor an opt-in framework combined with stronger negotiation rights for content creators, arguing that such steps would create a more equitable environment.
They stress that swift enforcement is critical. Courts have already ruled against it in separate monopoly cases involving search and digital advertising. Delaying action, they argue, risks irreversible harm to the open web and to the financial stability of news organizations that rely on referral traffic.