An artificial intelligence hallucination occurs when a generative model produces information that is not supported by its input, available evidence or established facts. The output can sound coherent and confident even when a diagnosis, citation, medication or other detail was invented.
Hallucinations are especially dangerous in clinical work because a fabricated detail can alter a medical record or decision. Grounding, constrained generation, source comparison, targeted evaluation and human review can reduce risk, but no single safeguard guarantees correctness.
Acronyms and aliases
AI hallucination acronymartificial intelligence hallucination variantconfident model error variant
Related terms
Frequently asked questions
Why do AI hallucinations look convincing?
Language models generate statistically plausible text, so style and fluency can remain convincing even when the underlying statement lacks evidence.
How can clinical AI hallucinations be detected?
Outputs can be compared with source records, current guidance and expert judgment using checks tailored to diagnoses, medications and decisions.
