Google Gemini 3 AI Overviews achieve approximately 90% accuracy rate

An analysis of Google's Gemini 3-powered AI Overviews indicates the feature generates accurate responses approximately 90% of the time, according to a report from the New York Times. The accuracy rate, while high in percentage terms, still means that across Google's estimated 5 trillion annual searches, tens of millions of AI-generated answers could contain errors every hour. The calculation highlights the scale of potential misinformation as search engines increasingly integrate large language models. Google introduced AI Overviews to provide summarized answers directly in search results, drawing from multiple web sources. The feature represents a significant shift from traditional search results that simply link to external websites. The accuracy assessment comes as technology companies face scrutiny over the reliability of AI-generated content. Regulators and researchers have raised concerns about hallucinations and factual errors in large language model outputs, particularly for sensitive queries related to health, finance, and current events.