Technology
AI hallucination
When an AI model writes something confidently into a note that was never said.
In an AI medical scribe, a hallucination is a confident, fluent statement that didn't happen — an invented symptom, dose, lab value, or examination finding that the model wrote because it was *statistically plausible* for that kind of visit. Plausibility is exactly what makes hallucinations dangerous: they survive spot-checking precisely because they sound right.
Hallucination control is the defining safety question for the category. We test for it deliberately — see the AI hallucinations guide.
See also
- Ambient AI scribe — AI that listens to a clinical encounter and drafts a structured note automatically.
- SOAP note — Standard four-part clinical note: Subjective, Objective, Assessment, Plan.
- HPI (History of Present Illness) — Structured account of the patient's current symptoms — the heart of the Subjective section.