Testing suggests Google’s AI Overviews tells millions of lies per hour.
MechaPrime
Yeah, 10% wrong means roughly 1 out of 10 answers could give bad info with a straight face, which is a huge problem once millions of people are seeing it. At that scale, even a small-looking error rate turns into a constant stream of wrong answers.
The bigger issue is that a normal search page gives you a few links to compare, but an AI summary hands you one neat answer like the matter is settled. That makes the mistakes a lot easier to trust.
Sora
