Diagnosis (general) · AI doctors and chatbots
AI scribes that write up your appointment: the first randomised trials
In November 2025 NEJM AI published two randomised trials of 'ambient' AI scribes, which listen to a consultation and draft the doctor's notes. In one trial, one of the two scribes modestly reduced time spent writing notes; in the other, AI reduced clinician exhaustion scores. Doctors reported occasional clinically significant inaccuracies, so notes still need checking.
How far away is this?
These tools are already used in many clinics, but how they're regulated varies and we haven't verified it anywhere. The best evidence is two short randomised trials with modest effects, so we rate it In clinical trials.
Where is it approved?
| Region | Regulator | Status | Notes |
|---|---|---|---|
| United States | FDA | Unknown | Whether scribe software counts as a medical device depends on what it does. We haven't verified any product's status. |
| European Union | CE mark / EMA | Unknown | |
| United Kingdom | MHRA | Unknown | |
| Australia | TGA | Unknown |
Ambient scribes are widely used, but their regulatory status varies by product and country, and we haven't verified it. All regions are marked Unknown. "Unknown" means we haven't verified it, not that it's unavailable. Approval in one country isn't approval in another.
What is it?
With your consent, an app records the conversation between you and your clinician and drafts the clinical note. The clinician reviews and edits it. The aim is more eye contact and less typing.
What's the evidence?
A UCLA three-arm randomised trial (238 outpatient physicians) compared Microsoft DAX, Nabla and usual care. Published in NEJM AI, 26 November 2025.[1]
Evidence stage: In clinical trials · #c1
In the published trial, Nabla cut time spent writing notes by 9.5% versus usual care; DAX showed no significant change. Both scribes showed possible improvements in burnout-related scores, which the authors say need confirming. Clinically significant inaccuracies were reported 'occasionally'. The trial ran for about two months at a single US academic health system and was funded by UCLA's Department of Medicine and others.[2]
Evidence stage: In clinical trials · #c2
A second randomised trial (66 practitioners, University of Wisconsin, 24 weeks) found ambient AI reduced work exhaustion scores and time spent on notes (about 20 minutes a day less), but did not significantly increase professional fulfilment. It was funded by UW Hospitals and Clinics and the US National Institutes of Health. Neither trial was designed to measure harm to patients from note errors.[3]
Evidence stage: In clinical trials · #c3
What does it mean for me?
Your clinician may ask to record your appointment so an AI can draft the notes. You can ask how the recording is stored and say no. If you can see your records, it's worth checking them for mistakes. If your notes or after-visit summary contain a wrong medicine, allergy or dose, tell your clinic promptly.
What could I ask my doctor?
- Is an AI recording or summarising this appointment?
- Where is the recording stored, and is it deleted?
- Can I opt out, and will that affect my care?
These are general starting points, not a recommendation to choose any particular test or treatment.
Change history
- : First published.
- : Updated after independent fact-check and clinical review
- : Review label corrected: this page was not clinically reviewed. Earlier labels saying 'independent fact-check' and 'clinical review' were wrong; checks so far were AI-assisted source checks only.
General information only, not medical advice. Talk to your doctor about your situation. AI-assisted source check. Not clinically reviewed.
References
- NEJM AI: Ambient AI Scribes in Clinical Practice: A Randomized Trial. doi.org. Peer-reviewed randomised trial. https://doi.org/10.1056/aioa2501000
- PubMed 41497288: UCLA trial abstract. pubmed.ncbi.nlm.nih.gov. Abstract. https://pubmed.ncbi.nlm.nih.gov/41497288/
- NEJM AI: pragmatic RCT of ambient AI and practitioner well-being. doi.org. Peer-reviewed randomised trial. https://doi.org/10.1056/aioa2500945
- PubMed 41625485: Wisconsin trial abstract. pubmed.ncbi.nlm.nih.gov. Abstract. https://pubmed.ncbi.nlm.nih.gov/41625485/
Publisher shown as the website hosting each source. Dates are given where the source label includes them.