9% of Google's AI Overviews are inaccurate

Source: Oumi, commissioned by The New York Times

Study: Accuracy analysis of Google AI Overviews

Key Findings:

• Google's AI Overviews are accurate 91% of the time using Gemini 3, up from 85% with Gemini 2. Google disputes the methodology
• At 5 trillion searches per year, that 9% error rate equals tens of millions of wrong answers every hour and hundreds of thousands every minute
• Google's own internal analysis found Gemini 3 produced incorrect information 28% of the time, though Google argues AI Overviews are more accurate because they draw on search results first
• Ungrounded responses, where AI cites websites that don't actually support its claims, jumped from 37% with Gemini 2 to 56% with Gemini 3, making it increasingly difficult for users to verify anything
• Only 8% of users double-check what AI tells them. Studies show users continued to follow AI guidance even when it gave the wrong answer nearly 80% of the time, a pattern researchers have dubbed "cognitive surrender"sites.
• Real errors caught in the analysis include: Google stating Hulk Hogan had died, citing a Bob Marley museum opening date off by a year, and cases where AI confidently cited sources that contradicted its own summary
• AI Overviews are also vulnerable to manipulation, blog posts have successfully misled the AI into presenting unqualified individuals as authorities in unrelated fields

Risks & Advantages

This isn't a Google story. It's a human cognition story...

The technology is wrong hundreds of thousands of times per minute, but that's not the most alarming finding. The most alarming finding is that 92% of users never check. Researchers have a name for it: cognitive surrender: the moment you stop verifying and simply accept what the machine tells you. That's the Human Edge risk in its most visible form. AI doesn't have to be malicious to erode your judgment. It just has to be convenient enough that you stop exercising it.

When AI answers come pre-packaged at the top of every search — authoritative in tone, fast, and confident even when wrong — the habit of independent verification quietly disappears. And habits, once lost at scale, are very hard to rebuild.

The Competitive Advantage:

The answer isn't to stop using Google. It's to stay cognitively in command while you do.

• Treat AI Overviews as a starting point, never an endpoint. If the answer matters, either professionally, medically, legally, or factually, click through. One source check is the difference between being informed and being misled.
• Notice when you've stopped verifying. That's the exact moment cognitive surrender begins. The habit of checking is a muscle. Use it or lose it.
• Ungrounded citations are the hidden danger. With 56% of AI Overview citations not actually supporting the claims they're attached to, confident sourcing is no longer a reliable signal of accuracy. Your judgment has to fill that gap.
• Independent judgment is now a competitive advantage. In a world where most people are accepting AI output at face value, the person who verifies, cross-references, and thinks critically stands out and stays irreplaceable.

The researchers called it cognitive surrender. The Human Edge calls it the risk nobody's building a strategy around... yet.