Google's AI Overviews feature provides AI-generated summaries above search results. While users cannot disable the feature, they can use the Web filter to view only standard links. The system generates answers by summarizing snippets from search results, which can lead to inaccuracies, outdated information, or mixed jurisdictions.
Research by Oumi for The New York Times in April 2026 evaluated 4,326 queries on Gemini 3, showing a 91% accuracy rate for factual questions, up from 85% with Gemini 2. However, 56% of these correct answers contained sources that did not fully support the claims. Google disputed the study, citing flaws in the benchmark and methodology.
A separate study by EBU and BBC in October 2025 found that 76% of Gemini app responses to news queries had significant issues, with 72% showing problems with sources. Additionally, a Pew Research Center study from July 2025 indicated that users click on AI Overview links in only 1% of visits, despite the feature appearing in 18% of searches.
Experts recommend verifying AI-generated data by checking primary sources, confirming dates, and comparing information against official government or institutional websites. For critical tasks like medical advice or legal strategies, AI summaries should not be used as a primary source of truth due to the risk of hallucinations and context errors.
