81% of physicians reported using AI professionally in 2026, up from 38% in 2023, according to the American Medical Association's Physician Survey on Augmented Intelligence. The average physician now names 2.3 AI use cases in daily work, up from 1.1 three years earlier, and the number carrying the most weight for practice administrators is smaller: 27% of physicians say they got zero formal AI training from any source.
The AMA fielded the survey between January 15 and February 2, 2026, drawing 1,692 responses, up from 1,183 in the 2024 wave and 1,081 in the 2023 wave. For independent practices and the billing companies that support them, the growth curve is not the interesting part. The gap between how fast physicians adopted AI and how little structure practices put around that adoption is the part that creates exposure.
How fast adoption actually moved
The jump from 38% to 81% in three survey waves means AI use crossed from a minority behavior to a majority one inside a single practice-management planning cycle. Average use cases per physician more than doubled over the same period, from 1.1 to 2.3, which points to physicians layering AI into more parts of a single workday rather than trying it once and stopping.
Confidence in the technology moved with adoption. 76% of physicians in the 2026 wave said AI improves their ability to care for patients, up from 65% in 2023. Use of AI for medical research summarization alone rose to 39% in 2026 from 13% in 2024, a jump the AMA data does not break out by specialty but that lines up with the broader adoption trend.
What administrators are being asked to govern
70% of physicians told the AMA they see AI as a tool for automating tasks that contribute to burnout, which puts documentation and administrative workflows squarely inside the adoption story rather than outside it. That is a practice-management decision as much as a clinical one, since burnout-reduction tools typically mean new software running against the EHR, new data flows, and new vendor contracts.
Governance expectations came through clearly in the same survey. 85% of physicians said they want a direct say in how AI tools get adopted inside their practice, and clear liability frameworks ranked as physicians' top regulatory priority. Practices that let individual physicians pick AI tools without a review process are adopting faster than they are governing.
The training gap by the numbers
92% of physicians in the 2026 survey said they want more AI education and training. 27% reported they had received none at all, from their employer, a specialty society, or any other source.
The demographic detail matters for staffing decisions. Respondents reported a median of 20 years in practice and 35 hours a week of direct patient care, and 38% work in group practices against 24% in hospital settings. More than a quarter of a workforce with two decades of median experience got no training before they started using AI at work.
Where physicians still draw a line
Adoption enthusiasm has limits that matter for policy-writing. 88% of physicians said they are concerned about skill erosion from relying on AI, a concern the AMA found runs higher among physicians earlier in their careers. 86% said data privacy protections are essential to their willingness to keep using AI tools, and about 40% describe their overall stance as equally excited and concerned rather than settled in either direction.
Patient-facing AI use draws a sharper line than administrative AI use. Nearly half of physicians surveyed strongly oppose patients using AI tools to interpret their own radiology or pathology results, even as most describe patient use of AI for general health questions as neutral or beneficial. A patient-communication policy on AI needs separate rules for those two situations, since physicians rate them differently.
Physicians are also unsure how much AI use their own patients are hiding. 29% told the AMA that no patients had disclosed using an AI chatbot to them, while 30% believed at least half of their patients already use one. 70% rated patient use of general-purpose AI for routine health questions as neutral or beneficial rather than harmful, a split between assumed disclosure and assumed use that has a direct bearing on intake forms and informed-consent language.
| Metric | 2023 | 2024 | 2026 |
|---|---|---|---|
| Physicians surveyed | 1,081 | 1,183 | 1,692 |
| Physicians using AI professionally | 38% | n/a | 81% |
| Average AI use cases per physician | 1.1 | n/a | 2.3 |
| Using AI for research summarization | n/a | 13% | 39% |
| Believe AI improves patient care | 65% | n/a | 76% |
Source: American Medical Association, Physician Survey on Augmented Intelligence, 2023-2026 waves.
What this means for practice operations
None of the AMA's figures describe billing or scheduling AI directly, since the survey asked about physicians' own AI use rather than practice-wide software decisions. But 81% adoption among the physicians a practice employs is itself an operational fact: it means most practices already have AI tools in use somewhere in the building, whether or not administration approved them.
The 27% with zero training and the 85% who want input on adoption decisions point to the same fix, a written AI-use policy that names which tools are approved, who trained on them, and who signs off before a new one gets added. That policy costs less to write now than the liability exposure the AMA's own respondents flagged as their top regulatory concern.
Sources
- American Medical Association, "AI usage among doctors doubles as confidence in technology grows," March 12, 2026
- American Medical Association, "Physician Survey on Augmented Intelligence," 2026
- HIT Consultant, "AMA Survey: 81% of Doctors Now Use AI, But Who Owns the Liability?," March 12, 2026
- The ASCO Post, "AMA Survey Finds Rapid Growth in Physician AI Adoption," March 2026
Talk to us
Fifteen minutes about your practice and the systems you run.