Google's AI Overviews, which debuted in 2024, has been scrutinized for its accuracy in providing information. A recent analysis by The New York Times, aided by the AI startup Oumi, found that while the AI's accuracy has improved to 90% following the Gemini 3 update, this still means it generates hundreds of thousands of incorrect answers every minute. The SimpleQA evaluation, which tests AI models with over 4,000 verifiable questions, revealed that 1 in 10 answers from AI Overviews is wrong, leading to concerns about the reliability of information users receive. The report highlighted specific instances where the AI provided incorrect information, such as conflicting dates regarding Bob Marley’s former home and erroneous statements about the Classical Music Hall of Fame. These findings underscore the challenges Google faces in ensuring the accuracy of its AI-driven search tools.