Hallucination

In AI, a hallucination is fluent but false or fabricated output from a model, not a human sensory experience. LLMs generate plausible text without a reliable built-in check that claims are true.

The gap between confident wording and factual accuracy matters for chatbots, RAG systems, and invented citations. Reducing hallucinations is a practical concern in alignment and product safety.