Chatbots Hallucinate at Least 15% of the Time — and Google Assistant Got Worse After AI Integration
New data puts hallucination rates between 15% and 38% across major chatbots, while users report that Google's AI-augmented assistant has regressed in basic reliability.
The hallucination problem isn't getting better fast enough. As @zerohedge reported in a viral post, recent analysis shows chatbots fabricate information at least 15% of the time, with ChatGPT's rate reportedly reaching 35%. These aren't edge cases or adversarial prompts — they're rates observed during normal usage patterns. For any enterprise deploying these models in customer-facing or decision-support roles, a one-in-three error rate is disqualifying.
Unlock the full briefing
Get every story in today's briefing, the full archive, and the daily AI intelligence brief.
All stories today
Full archive
Daily brief
Cancel anytime. Payments powered by Stripe.