New benchmark shows AI chatbots lose two‑thirds of correct answers when asked to self‑verify. Even GPT‑5.5, Claude 4.6 Sonnet and Gemini 3.5 Flash stumble on uncertainty. What does this mean for AI advice? Dive in for the full study. #GPT55 #Claude4_6 #AIUncertainty

🔗