Flags ungrounded confidence and hedge-word patterns in LLM responses. Spot when AI is guessing instead of verifying.
