Hallucination

Hallucination is when a language model generates output that is fluent and confident-sounding but factually wrong or completely fabricated. A model that invents a fake citation, an imaginary court case, or a non-existent API endpoint is hallucinating.

Hallucination is mostly orthogonal to AI detection — a piece of text can be obviously AI-generated and accurate, or beautifully human-feeling and full of made-up "facts." Detection tools tell you who (or what) wrote something, not whether what was written is true.

For trust & safety workflows, you usually want both: AI detection for provenance, and separate fact-checking for accuracy.