Google AI Overviews generate millions of inaccurate answers daily
An analysis by AI startup Oumi, reported by the New York Times, found Google's AI Overviews are accurate approximately 90% of the time, meaning tens of millions of erroneous answers are delivered daily given Google's scale. Google disputed the study's methodology, claiming it relies on a flawed benchmark test.
What happened
Google launched AI Overviews in 2024, giving artificial intelligence-generated answers prime placement at the top of search results. At the request of The New York Times, AI startup Oumi analyzed Google's AI Overviews using SimpleQA, a benchmark test with over 4,000 questions with verified answers. Oumi tested the system in October when Gemini 2 powered AI Overviews, finding 85% accuracy, and again in February after the upgrade to Gemini 3, finding 91% accuracy.
More than half of the accurate responses were "ungrounded," meaning they linked to websites that did not completely support the information provided. Facebook and Reddit were the second- and fourth-most-cited sources across 5,380 sources cited by Google's AI Overviews during the analysis, with Facebook cited more frequently in inaccurate answers than accurate ones.
Reported impact
With Google processing more than 5 trillion searches a year and AI Overviews achieving approximately 90% accuracy, analysis suggests Google provides tens of millions of erroneous answers. According to the study, on Google's massive scale, this could translate into thousands of inaccurate answers daily.
Index analysis
Google acknowledged that its AI Overviews can include errors. However, Google spokesperson Ned Adriance stated in response to the analysis that "This study has serious holes," arguing that Oumi's analysis was flawed because it relied on SimpleQA, a benchmark test built by OpenAI that itself contained incorrect information.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.