A hallucination is when an AI generates something that sounds correct but is actually wrong, invented or unsupported — a fake statistic, a non-existent book, a misremembered date. It happens because language models are built to produce plausible-sounding text, not to look up verified facts. Fluency and accuracy are separate things.
This is the single most important reason to review AI output rather than trust it blindly. For anything factual — names, numbers, quotes, legal or medical claims — verify against a reliable source before you rely on it. Using AI for drafting and phrasing, then checking the facts yourself, gives you the speed without the risk.
Related terms
Large Language Model (LLM) · Generative AI