Why is asking Claude "Are you sure about that?" an unreliable hallucination check?
Answer: Claude will often reconfirm incorrect answers with equal or greater confidence because it optimizes for plausibility, not accuracy
LLMs optimize for plausible continuation. Asking for confirmation prompts Claude to generate a confident reaffirmation — not an accuracy check.