What is an ambient AI scribe?

An ambient AI scribe is software that listens to a clinical visit, transcribes the conversation and drafts the clinical note. The clinician then reviews the draft, edits it and signs it. "Ambient" means the clinician does not have to narrate the note while the patient is in the room.

An ambient scribe listens to the visit and drafts the note from the conversation

A medical scribe was historically a person who followed the visit and wrote the chart. The AI version replaces that person with a recording. It uses automatic speech recognition to turn the conversation into text, then a language model sorts that text into a note with sections for the history, findings and plan.

Scribes are now built into large record systems as well as sold as separate tools. In February 2026, Healthcare Dive reported that Epic had rolled out an AI tool that drafts clinical notes ¹. Microsoft's quick start guide for its scribe in Epic says to open Epic Haiku or Canto on a mobile device, select a patient, and then select DAX Copilot at the bottom of the screen ².

The draft reaches the record only after the clinician signs it

The output is a draft, and the electronic health record keeps the note only once the clinician approves it. That review is the safeguard. A model writing from a conversation can leave out something that was said or add something that was not, so the signing clinician is responsible for checking the draft.

Trials test the same workflow. In the UCLA randomized trial, 238 outpatient physicians across 14 specialties were assigned to one of two AI scribes or to usual care ³. The trial compared named products, so its results describe specific software rather than scribes in general.

A scribe listens to conversation, while dictation captures what you choose to say

Dictation captures speech the clinician deliberately narrates. An ambient scribe captures the conversation between clinician and patient, which the clinician may not narrate at all. That difference decides which tool fits your work. Outpatient visits with a real two-way exchange suit a scribe. Procedures, radiology reports and letters, where the clinician narrates findings, suit dictation. Our guide to AI medical scribes compares the two, and the question on scribe versus dictation works through the choice.

Yap, which we make, is a dictation app, not an ambient scribe. It turns speech you choose to dictate into text on your Mac, Windows PC or phone. For the workflow of clinicians who dictate, see the guide to medical dictation.

Most drafts are usable, but the review decides what reaches the record

Most AI drafts are accurate enough to sign, but not all of them are. In a 2024 pilot at UC Davis, physicians reviewed 356 notes made with an ambient scribe and found 94.7 percent free of significant errors ⁴. The rest contained errors that the physicians sorted by type. The question on how accurate AI scribes are explains those error types and how they were counted.

It depends on the product, the visit and the consent rules

• Products differ. In the UCLA trial, physicians who used Nabla cut their time-in-note by 9.5 percent against the control group ⁵. Users of Microsoft DAX Copilot showed no significant change in time-in-note ⁶. A product that works in one clinic may not work in another.
• Consent can apply to every party. California's Penal Code 632 punishes a person who intentionally records a confidential communication without the consent of all parties ⁷. Check the rules for your state with the guide to recording consent laws.
• Vendors must give assurances. HIPAA lets a covered entity allow a business associate to handle protected health information on its behalf only if it obtains satisfactory assurance that the business associate will safeguard the information ⁸. The HIPAA guide for dictation explains what to check in a vendor's terms.

This is general information, not legal advice. Ask your compliance office or counsel before you record a visit.

What to do: check each draft against the conversation, not only the page

1. Read each draft against your memory of the visit before you sign it.
2. Check medications, doses, allergies and the plan first, since a missed change there matters most.
3. Ask the vendor which product and version you are testing, and how it measured errors.
4. Pilot the tool in your own clinic with a comparison period, so you can see the effect on your work.
5. Set a consent step for every recorded visit, and sign a business associate agreement before any patient audio is recorded.

For the wider subject, see our guide to AI medical scribes and our overview of dictation for work and study. Two related questions are how much time doctors spend on documentation and whether AI scribes reduce physician burnout.