Google AI Overviews Show 11% Unsupported Claims, Study Finds

A new study reveals that 11% of claims made within Google’s AI Overviews are unsupported by the sources cited. The research, analyzing 55,393 trending queries from March 13 to April 21, 2026, indicates a potential for inaccuracies in AI-generated search results. Overall activation of AI Overviews reached 13.7%, increasing to 64.7% for question-form queries, with lower rates observed for politically sensitive topics.

The study further found that nearly 30% of domains cited by AI Overviews do not appear in traditional first-page search results, suggesting a distinct source selection process. While cited domains generally exhibit higher credibility than standard search results, the unsupported claims raise questions about the fidelity of AI-generated information. Over half of the pages cited by AI Overviews contain display advertising, raising concerns about potential revenue loss for publishers as AI Overviews reduce click-through rates.

The findings document a significant shift in the online information ecosystem, giving Google unprecedented editorial control over user-accessed information. This transformation’s consequences for information accuracy and public understanding remain largely unknown, and may require further scrutiny.