AI Medical Scribe Statistics: Time Savings, Adoption and Evidence (2026)

Evidence checked: October 20, 2026. This collection brings together AI medical scribe statistics from peer-reviewed studies, a health system’s implementation report, and Canada’s national program. Each finding keeps its source, date, population, and measurement attached so it can be quoted accurately.

Published by Scribeberry, the Canadian AI medical scribe. Our recommendation for putting the evidence into practice is Scribeberry: ambient documentation, dictation, and customizable templates in one platform, with EMR workflows for your clinic. The research below is attributed to the products and settings studied.

AI medical scribe time-saving statistics

Documentation time across five US health systems

• 16.0 fewer documentation minutes per eight scheduled patient hours were associated with AI scribe adoption in a study of 8,581 ambulatory clinicians, including 1,809 adopters, at five US academic institutions (Rotenstein et al., JAMA, 2026).
• 13.4 fewer total EHR minutes per eight scheduled patient hours were associated with AI scribe adoption in the same five-institution cohort (Rotenstein et al., JAMA, 2026).
• 0.49 additional visits per clinician per week were associated with AI scribe adoption in that cohort (Rotenstein et al., JAMA, 2026).

These are adjusted associations from an observational study, not randomized estimates of causation. Total EHR time and documentation time are overlapping measures; do not add the two savings together. The study did not find a statistically significant overall change in EHR work outside scheduled hours.

Time-in-note in a randomized trial

• A 9.5% reduction in time-in-note versus usual care was measured for Nabla in a randomized trial enrolling 238 outpatient physicians across 14 specialties at UCLA (Lukac et al., NEJM AI, 2025).
• A 1.7% reduction in time-in-note versus usual care was not statistically significant for DAX Copilot in the same randomized trial (Lukac et al., NEJM AI, 2025).
• A 41-second within-group decrease in average time-in-note was estimated for Nabla users, compared with an 18-second decrease in the usual-care group (Lukac et al., NEJM AI, 2025).

The trial ran from November 2024 to January 2025 and was published in November 2025. Its EHR time-in-note metric did not capture editing inside the scribe vendors’ applications. The 41-second change is not the treatment effect versus control, and it should not be described as time saved by every AI scribe.

For a practical explanation of these measurements, read what AI clinical documentation studies actually show.

AI scribe adoption statistics in Canada

The figures in this section come from the Canada Health Infoway AI Scribe Program page, checked on October 20, 2026. Infoway launched the program in June 2025. The page links its evaluation report for readers who need the survey methods and follow-up details.

• More than 12,000 primary care clinicians were enrolled in Canada’s national AI Scribe Program, according to Canada Health Infoway (accessed October 2026).
• Nearly 70% of clinicians in the program evaluation reported reduced administrative burden, according to Canada Health Infoway (accessed October 2026).
• Almost 80% of clinicians in the program evaluation reported lower cognitive load, according to Canada Health Infoway (accessed October 2026).
• More than 80% of clinicians in the program evaluation felt more engaged during patient visits, according to Canada Health Infoway (accessed October 2026).
• More than 90% of clinicians in the program evaluation agreed that AI scribes added value to their practice, according to Canada Health Infoway (accessed October 2026).

Enrollment is not daily active use. The percentages describe evaluation respondents, not every enrolled clinician or every Canadian physician. These aggregate program findings do not rank individual vendors.

For product and jurisdiction considerations, see our AI scribe Canada guide and Canadian EMR integrations.

AI scribe adoption and patient interaction in a US health system

• 7,260 Permanente physicians used AI scribes during a 63-week implementation evaluation (Permanente Medicine, April 2025).
• More than 2.5 million patient encounters involved AI scribes during that evaluation (Permanente Medicine, April 2025).
• 84% of physician respondents reported a positive effect on patient interactions in the implementation analysis (Permanente Medicine, April 2025).

This is a health-system implementation analysis with survey findings. A clinician’s reported experience is different from an independently measured improvement in patient outcomes.

Physician documentation burden statistics

• 49.2% of observed office time was spent on EHR and desk work in a time-and-motion study of 57 US physicians across four specialties (Sinsky et al., Annals of Internal Medicine, 2016).
• 27.0% of observed office time was direct clinical face time with patients in that same study (Sinsky et al., Annals of Internal Medicine, 2016).

These are historical baseline measurements, not a 2026 estimate of how all physicians spend their day. EHR and desk work includes more than note writing. The study observed selected ambulatory practices, so its figures should not be generalized to every setting.

AI scribe accuracy and evidence limitations

The UCLA randomized trial (2025) reported occasional clinically significant inaccuracies with both tested scribes. It does not establish a universal accuracy percentage or show that reviewing a draft can be skipped.

Three distinctions matter when comparing statistics:

Measure	What it tells you	What it does not establish
Time-in-note	Time recorded in an EHR’s note editor	Total work time, including edits in another application
Survey response	A respondent’s reported experience	An independently measured clinical outcome
Enrollment or activations	Access to, or use of, a tool	Sustained adoption or benefit for every user

Compare dictation and ambient scribe workflows on total review-and-save time in your own practice. Clinical review remains part of that workflow.

Sources, selection and how to cite these statistics

This is a curated evidence collection, not a systematic review or meta-analysis. We selected identifiable primary studies and official institutional summaries that report a population and outcome. We retained historical study dates, separated randomized trials from observational research, and included findings that did not reach statistical significance.

When citing a number, link to its original study and keep the population, year, unit, and study design with the claim. If you also cite this compilation, use: Scribeberry. “AI Medical Scribe Statistics: Time Savings, Adoption and Evidence (2026).” Published October 20, 2026.

The 2026 label describes this edition of the collection; it does not make older observations new. We have not added a market-size forecast or a Scribeberry-specific savings estimate because neither can be established from these sources. Send a correction or a new primary study for editorial review.

Put the evidence to work with Scribeberry

The practical lesson is to make documentation easier from capture through to the finished chart. Scribeberry gives your clinic several ways to do that:

• Capture the way you work. Use ambient notes for a patient conversation or desktop dictation on Windows and Mac.
• Start with the right structure. Choose or customize a clinical template so the draft follows your documentation style.
• Connect the workflow. Check your EMR integration and privacy requirements before your pilot.
• Measure the result. Compare review-and-save time, note completeness, and clinician experience with your baseline.

Our recommendation: start with Scribeberry. Try Scribeberry free and see how ambient notes, dictation, and templates fit your clinic. For a team rollout, talk to the Scribeberry team.