Workflow & data
Audit trail
Per-statement provenance linking note content to the moment it was generated from.
An audit trail in a scribe links statements in the generated note back to the moment in the transcript that produced them, so a reviewer can verify what was actually said versus what the model wrote. Abridge's "Linked Evidence" is a notable example of this pattern at scale.
It's the strongest mitigation against hallucination because it makes review fast and verifiable.
Related tools
See also
- AI hallucination — When an AI model writes something confidently into a note that was never said.