Key takeaways
- 81% of physicians in the AMA's 2026 survey reported using AI tools professionally, more than double the 2023 rate.
- The most common specific uses are administrative: documentation, drafting portal replies and translation.
- Seven in 10 physicians see AI as a way to automate burnout-causing tasks, but privacy remains a top concern.
- Practices should turn the national picture into a short internal survey and a written AI policy.
The headline number
Physician use of AI has moved from early adopters to the majority in three years. In the American Medical Association's 2026 survey, 81% of physicians reported using AI tools professionally, more than double the rate in the AMA's first health AI poll in 2023[1]. Belief in the benefit has grown too: more than three-quarters of physicians in 2026 said AI improves their ability to care for patients, up from 65% in 2023[1].
Use does not mean deep use. A physician who asks a general AI tool to summarize a journal article counts the same as one whose notes are drafted by an ambient scribe in every visit. The more useful numbers are the specific uses, because they show where AI is entering daily work and where practice operations will feel it.
What physicians use it for
The specific uses in the AMA survey are mostly administrative. 28% of physicians use AI for documentation of billing codes, medical charts or visit notes, 19% to generate draft responses to patient portal messages, and 18% for translation services[1].
Each of those touches the front office. Documentation tools change how quickly notes are closed and claims can go out. Draft portal replies change who reviews patient messages and how fast they are answered. Translation tools change how staff communicate with patients who prefer another language, and bring their own compliance questions, covered below.
Excited and worried at the same time
The survey shows enthusiasm with conditions. Seven in 10 physicians see AI as a tool to automate tasks that contribute to work-related burnout[1]. About 40% said they are equally excited and concerned, with patient privacy and the integrity of the patient-physician relationship as top concerns[1]. Physicians told the AMA that AI is likely to help with work efficiency, diagnostic ability, cognitive overload, clinical outcomes, patient convenience, and stress and burnout[1].
- Enthusiasm centers on removing clerical work, not on replacing clinical judgment.
- Concern centers on privacy and on what patients experience.
- Both point to the same practical rule: use AI where it removes paperwork, and keep patients' trust at the center of every deployment.
The burden AI is being asked to lift
The interest in administrative AI makes sense against the workload. In one widely cited study, primary care physicians spent 5.9 hours of an 11.4-hour workday in the EHR, including 1.4 hours after clinic hours[2]. Burnout has eased but remains high: in 2025, 41.9% of physicians reported at least one symptom of burnout, down from 43.2% in 2024 and 48.2% in 2023[3].
For practice managers, that context matters when physicians propose new tools. The question is not whether AI is interesting but which hours it gives back, to whom, and what new review work it creates in exchange.
Practices are adopting faster than they are governing
Organizational data tells the same story from the practice side. AI played some role in patient visits at 83% of medical groups in August 2026, up from 71% a year earlier[4]. Yet only 20% of medical groups had AI governance or a formal AI-use policy in January 2026, with 22% developing one and 56% having none[5]. In many practices, physicians are choosing and using tools faster than the practice is writing rules for them.
Compliance points the survey raises
Two of the most common uses carry specific federal obligations for covered health entities. Translation is one. Under Section 1557 rules, machine translation of critical or technical patient content must be reviewed by a qualified human translator[6]. A physician who uses AI to translate discharge instructions or consent language needs a review step, not just a good tool.
Decision support is the other. Covered entities must make reasonable efforts to identify and mitigate discrimination risks from patient care decision support tools they use[7]. That applies to AI tools that influence clinical decisions, and it is a reason to keep an inventory of which tools are in use and for what.
| Common use | Operational effect | What to put in place |
|---|---|---|
| Documentation and notes | Faster note closure; different review habits | Clinician review before signing; audit of coding suggestions |
| Draft portal replies | Faster message turnaround | Clear rule on who reviews and sends; no unsent drafts left in queues |
| Translation | Easier communication with patients who prefer another language | Qualified human review of critical content; access to interpreters |
| Decision support | Suggestions at the point of care | Inventory of tools; review of discrimination risk |
Reading the survey carefully
- Self-reported use: physicians report what they think counts as AI, and definitions vary.
- Professional use is broad: it includes research and learning, not only patient care.
- National averages hide specialty and setting differences; a small rural practice may look nothing like the average.
- Belief in benefit is not measured benefit; it tells you about appetite, not outcomes.
- Year-to-year comparisons depend on the same questions being asked the same way.
None of this makes the survey less useful. It makes it a starting point for asking your own clinicians the same questions.
Patients are using AI too
Physicians are not the only ones changing habits. About a third of US adults, 32%, are turning to AI for health information and advice, according to a KFF poll in early 2026[8]. The same poll found 77% of the public concerned about the privacy of personal medical information provided to AI tools[8]. Patients will increasingly arrive having asked an AI tool about their symptoms, and some will ask whether your practice uses AI with their records. Staff should have a short, honest answer ready: which tools the practice uses, for what, and how patient information is protected.
What it means for the front office
- Portal messages may be answered faster once physicians use drafting tools, which can raise patient expectations for every channel, including the phone.
- Notes closed sooner means charges and referrals can move sooner; billing and referral staff may see work arrive in bigger, earlier batches.
- Translation tools change how bilingual staff spend time, but do not remove the need for qualified interpreters on calls and at visits.
- Staff will be asked about AI by patients and should know the practice's policy in plain words.
Ask physicians to tell the practice manager before adopting any new AI tool, so the front office can plan for the knock-on effects instead of discovering them.
Questions for your next partners' meeting
- Which AI tools are physicians using today, and has each been approved?
- Which tools receive patient information, and are agreements in place for each?
- Who reviews AI drafts and translations before they reach patients?
- Which administrative task would the physicians most like automated next, and how would we measure success?
- When will we have a written AI policy, and who owns it?
Turn it into an internal survey
Ask five questions
Which AI tools do you use for work? For which tasks? With what patient information? What would you most like automated? What worries you?
Include staff, not just physicians
Front office, billing and clinical staff use tools too, sometimes without anyone knowing.
Build the inventory
List every tool, what it does, what data it receives and whether a business associate agreement is in place where needed.
Close the gaps
Stop uses that send patient information to tools without agreements, and add review steps for translation and drafts.
Write the one-page policy
Approved tools, approved tasks, data rules, review rules and how to propose a new tool.
Repeat yearly
Run the same survey each year so you can see your own trend, not only the national one.
Example: a fictional six-physician internal medicine group ran this survey and found that four physicians used AI for portal reply drafts, two of them through a personal account with no agreement in place. The group moved everyone to the drafting tool inside its EHR, added a rule that drafts are reviewed and sent by the physician or a designated nurse, and listed the tool in its new AI policy.
Frequently asked questions
What share of physicians use AI?
In the AMA's 2026 survey, 81% of physicians reported using AI tools professionally, more than double the rate in the AMA's first health AI poll in 2023.
What do physicians use AI for most?
Among the specific uses reported, documentation of codes, charts or notes was most common, followed by drafting patient portal replies and translation.
What worries physicians about AI?
Patient privacy and the integrity of the patient-physician relationship were the top concerns among physicians who said they were equally excited and concerned.
Are there rules on using AI for translation?
Yes. Under Section 1557 rules, machine translation of critical or technical patient content must be reviewed by a qualified human translator.
References
American Medical Association (2026). More than 80% of physicians use AI professionally: AMA survey (opens in a new tab) โ
Annals of Family Medicine (2017). Tethered to the EHR: Primary Care Physician Workload Assessment Using EHR Event Log Data and Time-Motion Observations (opens in a new tab) โ
American Medical Association (2025). AMA: Physician burnout rates are falling, specialty gaps remain (press release) (opens in a new tab) โ
Medical Group Management Association (MGMA) (2026). As AI use expands, medical practice leaders shift attention to measurement (opens in a new tab) โ
Medical Group Management Association (MGMA) (2026). AI governance in medical group practices: Rules for the humans in the loop (opens in a new tab) โ
Legal Information Institute, Cornell Law School (current). 45 CFR ยง 92.201 - Meaningful access for individuals with limited English proficiency (opens in a new tab) โ
Legal Information Institute, Cornell Law School (current). 45 CFR ยง 92.210 - Nondiscrimination in the use of patient care decision support tools (opens in a new tab) โ
KFF (2026). KFF Tracking Poll on Health Information and Trust: Use of AI For Health Information and Advice (opens in a new tab) โ