Out of Claude and GPT, only Claude asks me questions if it does not fully "know" something, be it the matter at hand or my intentions. I think this is a good way to avoid "hallucinations". GPT just keeps spouting whatever it can.