A review finds AI scribes lose the patient narrative

Lucas Seuren of the University of Edinburgh’s Centre for Biomedicine, Self and Society published a review of 27 studies in BMJ Digital Health & AI on 3 September 2026 (DOI 10.1136/bmjdh-2026-000091), examining what ambient AI scribes do to a consultation.

The findings divide in two. On the record itself, scribes miss non-verbal communication — facial expressions, gestures, emotional state — and their summaries prioritise clinical data over the patient’s account of their own illness. On behaviour, patients who know a session is being recorded by AI may hesitate to raise substance use, domestic violence or mental health concerns, which are precisely the disclosures a documentation tool has no way to recover.

The clinician effect cuts both ways. Offloading note-taking frees attention for the conversation, but the review reports reduced memory recall and slower skill development, with some clinicians not recognising their own notes or not remembering patients at a follow-up visit.

“Many clinicians are excited about ambient AI scribes because they promise to cut down on paperwork,” Seuren said. “But the experiences of patients are poorly considered, and there are real risks that patients’ stories are lost.”

This is a review of existing literature, not new trial data. Its subject is a tool already in routine use.