Google AI Overviews Generate Unverified Claims in 11% of Searches

Google’s AI Overviews, launched in 1998, have transformed how users interact with search results. Unlike traditional ranked lists, AI Overviews now provide AI-generated summaries for billions of searches, giving Google unprecedented editorial control. A forthcoming study, to be presented at the October 2026 ACM Internet Measurement Conference, examined AI Overview activation patterns, source selection, claim accuracy, and potential financial impacts on publishers.

The research, conducted by Jacob Montgomery, Umar Iqbal, and Haofei Xu, analyzed 55,393 Google searches over 40 days in 2026. It found that AI Overviews appeared in 13.7% of all queries, with question-form searches triggering overviews 65% of the time. Longer searches and specific topics like hobbies and science were more likely to generate AI summaries, while political and legal queries were less likely.

The study also revealed that 11% of AI Overview claims were either unsupported or contradicted by their cited sources. While the overviews relied on credible sources, the researchers noted that 29.8% of cited domains did not appear in the first-page results, indicating a distinct source-selection process. Only 41.9% of overviews were fully grounded in their cited text, highlighting the need for users to verify information even from seemingly trustworthy citations.

The financial implications for publishers were also concerning. Over half of the cited webpages displayed advertising, and 2.2% of search pages with AI Overviews included Google-sponsored ads. This arrangement could preserve Google’s ad revenue while potentially reducing traffic and revenue for content publishers. The authors warned that without licensing or revenue-sharing arrangements, publishers reliant on AI Overviews may struggle to survive.