Should doctors use an AI scribe or dictation?

Choose by the kind of work, not by the tool. An ambient AI scribe records a conversation and drafts the note. Dictation takes what you say and turns it into text. Both produce a draft that the clinician must review and sign. For most doctors the honest answer is both: a scribe for visits where a dialogue matters, and dictation for the narrated parts of the record. For the full picture, see our guide to AI medical scribes.

Use a scribe for conversational visits and dictation for narrated work

A scribe earns its place when the note depends on what the patient and clinician said to each other. A follow-up visit, a medication review or a new complaint is a conversation, and a scribe can capture it without the clinician narrating. What an ambient AI scribe is explains the workflow.

Dictation fits work that the clinician already narrates. A procedure note, a radiology report, a discharge summary or a letter to a referring doctor is spoken as a monologue. A recording of the whole consultation can add material that a letter does not need. The medical dictation guide covers the workflow in more detail.

No large trial compares the two head to head, so the time evidence is indirect

The randomized trials of scribes compare a scribe with usual care, not with dictation. At UCLA, one product cut time-in-note by 9.5 percent against the control group ¹. The other product showed no significant change ². The savings depend on the product, which is why a vendor's claim is not enough.

The only direct comparison we found is a small pilot in palliative medicine. Two residents used an ambient scribe and also dictated the same visits. For one resident, documentation time was significantly shorter with the scribe. For the other, there was no significant difference ³. The authors say the study is too small to settle the question: "Given the small sample size and short study duration, this study is insufficiently powered to draw conclusions with respect to AI scribe benefits in real-world contexts." ⁴

So the time evidence does not show that one method always wins. It shows that a scribe can save time for some products and not others, and that dictation's time cost has not been measured against a scribe in a large trial.

Both methods leave errors, and the review step decides whether they reach the record

Errors matter more than minutes, because a note that is quick but wrong is not useful.

For scribes, the UC Davis pilot found that physicians rated 94.7 percent of the 356 reviewed notes as free from significant errors ⁵. The remaining notes were not all minor. In 19 notes, or 5.3 percent, the error was one that physicians rated as posing serious or imminent risk if it was not corrected ⁶. These counts describe a sample of the notes physicians reviewed, not an audit of every note.

For dictation, a 2018 analysis of dictated clinical notes found that 96.3 percent of speech-recognition drafts contained at least one error ⁷. Those notes were dictated in 2016, and the draft errors were measured differently from the scribe audit, so the two rates are not directly comparable. The same study found that after review, the notes physicians signed had an error rate of 0.3 percent of words ⁸.

The lesson is the same for both. A draft is not a signed note. Whatever method you use, the review before signing is the control that matters, and it takes time you should count.

It depends on your visits, your consent rules and your review time

• Mostly conversational visits. A scribe is designed for this work. Test it on your own visits before you rely on the time saving, since the trials differ by product.
• Mostly procedures, reports and letters. Dictation is the direct tool. A scribe would record a conversation you may not need in the note.
• A mix. Many clinicians use a scribe for visits and dictation for letters and procedure notes. Nothing in the evidence forbids this, but each tool needs its own consent and review steps.
• Recording rules. Some states require every party to a conversation to consent to a recording. Check your state's rules and the recording consent laws guide before any visit is recorded. The question on whether patients must consent to AI scribes covers the consent step.
• Vendor terms. A vendor that handles patient audio needs a contract that protects the data. The HIPAA guide for dictation lists what to ask.

What to do before you choose

1. List your recent notes by type. Count visits, procedures, reports and letters over a typical month. The split tells you which tool carries the load.
2. Time both methods in your own workflow. Use a comparison period, as the Wisconsin, Duke and UCLA trials did, and include review and correction time.
3. Ask the vendor how it measured errors. Ask whether drafts were audited against the recording, and which product version was tested.
4. Set a consent step for every recorded visit. Decide what you say to the patient and record it in the chart.
5. Keep the review step for every note, whichever tool you use. Do not sign a draft you have not read against the visit or your own words.

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, which fits the narrated work above. The medical dictation guide compares the two approaches in more depth.

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

Related guides and questions

• Up: dictation for work and study.
• Guide: AI medical scribes and ambient clinical documentation.
• Questions: How accurate are AI medical scribes?, Do AI scribes reduce physician burnout?, How much time do doctors spend on documentation?.
• Terms: medical scribe, electronic health record, word error rate.