Statistics

AI scribe adoption statistics

A running tally of how far ambient AI documentation has actually spread in healthcare — the deployments, the measured effects on clinician time and burnout, and the accuracy questions that stay open. As of July 2026.

By Jonas WeirReviewed by Jonas Weir · clinical reviewUpdated

The short version

  • The largest reported deployment, at The Permanente Medical Group, passed over 2.5 million ambient AI scribe uses in its first year — after 3,442 physicians used it across 303,266 encounters in the first ten weeks alone.
  • Across six health systems, clinician burnout fell from 51.9% to 38.8% after thirty days with an ambient scribe — about 74% lower odds of burnout.
  • Time savings in large real-world studies are real but modest: on the order of 13 to 16 minutes of EHR and documentation time per clinician per day.
  • A study of 1.2 million encounters found adopters generated 1.81 more work RVUs per week than non-adopters.
  • Accuracy is the open question: overall error rates run low, but ambient notes carry distinctive failure modes — hallucinated content and clinically important omissions.

Ambient AI scribes — tools that listen to a clinical encounter and draft the note — moved from demo to daily workflow faster than almost any technology in recent healthcare history. This page gathers what the published record actually shows about how far that spread has gone, and what it has produced. Every figure is dated and tied to a numbered source below. As of July 2026.

What counts as "adoption"?

The single hardest thing about adoption statistics is that no two studies mean the same thing by use. One counts every physician who was offered the tool. Another counts anyone who used it once. A third counts only clinicians who use it for more than half their visits. A headline percentage is close to meaningless until you know which of those it measures — so each number here is labelled with the denominator it came from.

The scale of deployment

The largest and best-documented deployment is at The Permanente Medical Group in Northern California. In the first ten weeks after launch, 3,442 physicians used the ambient AI scribe across 303,266 patient encounters, spanning a wide range of specialties and locations 1. One year in, the same group reported that the tool had passed over 2.5 million uses and had moved into a formal quality-assurance program for documentation at that scale 2.

Those are the numbers that anchor most conversations about adoption. They are unusually large, and they come from an integrated system with a single EHR and a coordinated rollout — conditions most organizations do not share. Treat them as the ceiling of what a well-resourced deployment looks like, not the median.

What does the time-and-burnout evidence show?

The most-cited benefit is relief from documentation load. A quality-improvement study across six academic and community health systems found that, after thirty days with an ambient scribe, the proportion of clinicians reporting burnout fell from 51.9% to 38.8% — roughly 74% lower odds of burnout, alongside improvements in cognitive task load and after-hours documentation 3.

Larger controlled work tempers the size of the effect. A study of roughly 1,800 clinicians published in JAMA measured about 13 fewer minutes per day in the EHR (a 3% relative decrease) and 16 fewer minutes per day on documentation (a 10% relative decrease), plus about half an additional patient visit per week — while noting that only 32% of users adopted the tool for more than half their visits 5. The relief is real; the depth of use is uneven.

On the productivity side, a study of 1,202,734 encounters across 1,565 physicians found that adopters generated 1.81 more work RVUs per week than non-adopters over the study window 4. Adoption in that sample sat at about 45% of physicians — a reminder that even where the tool is available, fewer than half may take it up.

How accurate are AI scribe notes?

None of the adoption numbers matter if the notes are wrong. A real-world evidence synthesis in JMIR AI found that modern ambient scribes report low overall error rates, but carry a distinctive set of failure modes: hallucinated content the clinician never said, clinically important omissions, and misattribution 6. These are different from the transcription errors of older dictation tools, and they are the reason every published deployment keeps a clinician in the loop to review each note before it is signed.

How to read these numbers

Four cautions travel with every figure on this page. Deployments that report the biggest effects tend to be the best-resourced, so selection is at work. Burnout and satisfaction gains lean on self-report over short windows. Time savings measured in controlled designs are consistently smaller than the ones users describe. And the tools themselves differ enough that a result for one vendor does not transfer cleanly to another.

We revisit this page on a ninety-day cycle, and whenever a new large deployment or controlled study lands. If you are weighing an ambient scribe for your own setting, the deeper reads are in our guides on what the evidence actually shows and the documentation-time and burnout evidence.

Questions & answers

  • How widely are AI scribes actually used in healthcare?

    The clearest published picture comes from The Permanente Medical Group, which reported over 2.5 million ambient AI scribe uses in the first year of its deployment. Most other health systems report pilots or partial rollouts rather than system-wide daily use, and studies measure adoption differently — from being offered the tool, to using it once, to using it for the majority of visits. Read each figure for which of those it means.

  • Do AI scribes reduce physician burnout?

    A study across six health systems found that the share of clinicians experiencing burnout fell from 51.9% to 38.8% after thirty days with an ambient scribe. That is a meaningful signal, but it comes from a short-window quality-improvement design that relies on self-report, so it is best read alongside the more modest time-savings measured in larger controlled studies.

  • How accurate are the notes AI scribes produce?

    Overall reported error rates are low, but the errors that do occur have a distinctive shape — hallucinated content the clinician never said, and omissions of clinically important detail. This is why every published deployment pairs the tool with clinician review of each note before it enters the record.

Sources

  1. Tierney AA, Gayre G, Hoberman B, et al. Ambient Artificial Intelligence Scribes to Alleviate the Burden of Clinical Documentation. NEJM Catalyst Innovations in Care Delivery. 2024;5(3). doi.org/10.1056/CAT.23.0404
  2. The Permanente Medical Group. Ambient Artificial Intelligence Scribes: Learnings after 1 Year and over 2.5 Million Uses. NEJM Catalyst Innovations in Care Delivery. 2025. doi.org/10.1056/CAT.25.0040
  3. Use of Ambient AI Scribes to Reduce Administrative Burden and Professional Burnout. JAMA Network Open. 2025;8(10):e2534976. doi.org/10.1001/jamanetworkopen.2025.34976
  4. Ambient Artificial Intelligence Scribes and Physician Financial Productivity. JAMA Network Open. 2026;9(1):e2553233. doi.org/10.1001/jamanetworkopen.2025.53233
  5. Changes in Clinician Time Expenditure and Visit Quantity With Adoption of Artificial Intelligence–Powered Scribes. JAMA. 2026. doi.org/10.1001/jama.2026.2253
  6. Real-World Evidence Synthesis of Digital Scribes Using Ambient Listening and Generative Artificial Intelligence for Clinician Documentation Workflows: Rapid Review. JMIR AI. 2025;4:e76743. doi.org/10.2196/76743