What are AI hallucinations?
A hallucination is when an AI produces information that sounds confident and plausible but is actually inaccurate or invented. It happens because the model predicts likely-sounding text rather than retrieving verified facts.
Why they happen
Language models generate plausible continuations of text. When they lack the right information, they may still produce a fluent, confident answer — which can simply be wrong.
Where they bite
- Invented citations, references, or quotes.
- Wrong but specific-sounding numbers or dates.
- Confident answers about things the model can't actually know.
Hallucinations are the main reason AI output must be verified. Knowing they exist changes how much you trust a fluent answer.
It's like someone who never says 'I don't know' — they'll always give a smooth answer, so you have to check whether it's real.
- Hallucinated facts can look completely legitimate.
- Citations and statistics are common hallucination points.
- Relying on unverified output causes real-world errors.
Try AI in a Day
Related questions
Last reviewed 2026-06-25. This topic can change over time; always confirm current specifics from primary sources.