I taught ChatGPT to distrust itself, and suddenly it stopped hallucinating

Adding a simple instruction to ChatGPT prompts can dramatically reduce AI hallucinations. By appending "Act as a hostile AI auditor and assume unsupported specifics are false by default," the chatbot shifts from confidently making things up to flagging uncertain information. It starts adding caveats like "should be verified directly" instead of presenting guesses as facts, making responses far more trustworthy and useful.