This article explains five common “prompt traps” that can produce confident but unreliable outputs. Like hidden assumptions, demanding certainty, missing context, asking AI to “verify” facts it can’t check, or handing AI decisions that should stay with people. It offers safer prompt alternatives and a quick pre-submit checklist to help users get more accurate, responsible results in tools such as ChatGPT Edu and Copilot.