Analysis Shows Google’s AI Overviews Are Misinforming Users at Significant Scale

A new analysis by AI startup Oumi reveals that Google’s AI Overviews are producing misinformation at massive scale, despite being accurate around 91 percent of the time. With Google handling roughly five trillion searches per year, this error rate translates into tens of millions of incorrect answers every hour, raising concerns about users’ tendency to trust AI without verification. Tests show that while Gemini 3 improved factual accuracy over Gemini 2, it also produced more “ungrounded” responses that cite sources not supporting the claims. Google disputes the findings, but its own internal tests indicate significant accuracy issues, highlighting a growing crisis in AI‑generated search results.

Read the full story on Futurism →