How Doctors Actually Use AI Today

Survey data suggest doctors mostly use AI for reading and writing, not for making diagnoses. In the American Medical Association's 2026 survey, 81% of physicians said they use AI professionally. The most common tasks were summarizing medical research and drafting notes and paperwork. Diagnostic support ranked lower. Many hospitals use predictive software built into their record systems. Radiology accounts for most of the AI devices authorized by US regulators.

How many doctors use AI, and for what?

The AMA published its survey of nearly 1,700 physicians in March 2026. The 81% rate was more than double the level in its first poll in 2023. Physicians said they were using AI for:

• Summaries of medical research and standards of care: 39%
• Discharge instructions, care plans or progress notes: 30%
• Billing codes, medical charts or visit notes: 28%
• Chart summaries: 28%
• Draft replies to patient portal messages: 19%
• Translation: 18%
• Assistive diagnosis: 17%

More than three-quarters said AI improves their ability to care for patients. Seven in ten saw it as a way to automate tasks that contribute to burnout.

What is an AI medical scribe, and does it work?

An AI scribe, also called ambient documentation, drafts the clinical note from the conversation during a patient visit. The clinician then reviews it.

A UCLA Health trial published in NEJM AI in 2025 involved 238 physicians across 14 specialties. Each was randomly assigned to one of two commercial scribes, Microsoft DAX or Nabla, or to their usual practice. UCLA Health said Nabla users cut the average time spent writing each note by an estimated 41 seconds, to 3 minutes 49 seconds. It said the control group's time fell by 18 seconds. DAX users saw modest savings that were not statistically significant. Burnout scores improved by about 7% in both scribe groups relative to the control group. Physicians reported that AI-drafted notes occasionally contained clinically significant inaccuracies, most often omissions or pronoun errors. The two-month study took place at one academic medical center.

The University of Wisconsin said in December 2025 that a randomized trial of the Abridge scribe at UW Health cut documentation time by 30 minutes per provider per day. It said that trial was also published in NEJM AI.

A study in JAMA Network Open in October 2025 surveyed 263 clinicians at six health systems before and after 30 days with a scribe. The share reporting burnout fell from 51.9% to 38.8%. But the study had no control group and relied on self-reports.

How are hospitals and radiology departments using AI?

The FDA says it has authorized more than 1,600 AI-enabled medical devices as of September 2026. Trade publication MedTech Dive, which tracks the agency's list, reports that most are in radiology. A June 2026 Congressional Research Service report says most reached the market through the 510(k) pathway. That pathway requires showing substantial equivalence to a legally marketed device. As of early September 2026, the FDA had not authorized a device built on generative AI, the technology behind chatbots, MedTech Dive reported. General-purpose chatbots sit outside that system. The safety nonprofit ECRI notes they are not regulated as medical devices.

Hospitals also use predictive models that produce risk scores. A September 2025 federal data brief from the Office of the Assistant Secretary for Technology Policy looked at how hospitals use them. It found that 71% of non-federal acute care hospitals used predictive AI integrated with their electronic health record in 2024, up from 66% in 2023. It found hospitals used it most often to predict health trajectories or risks for inpatients. The fastest-growing uses were administrative. Among hospitals using predictive AI, billing rose from 36% to 61% and scheduling from 51% to 67%.

Adoption is uneven. In 2024, 96% of large hospitals used predictive AI, compared with 59% of small ones. The rate was 86% for hospitals in multi-hospital systems, compared with 37% for independent hospitals.

What worries doctors about AI?

Skill loss is a widespread concern. In the AMA survey, 88% of physicians reported it at least to some degree. An observational study in The Lancet Gastroenterology & Hepatology looked at four endoscopy centers in Poland. It measured how often doctors detected adenomas, a type of polyp, in colonoscopies performed without AI. After AI assistance was introduced, that rate fell from 28.4% to 22.4%, the study found. The study was retrospective and cannot prove that AI caused the decline.

Accountability is another. The AMA reported that physicians ranked clear liability frameworks highest among the regulatory steps needed to build trust. It also reported that nearly half strongly opposed patients using AI to interpret radiology or pathology results.

What to watch

Three things stand out as of October 2026:

• Longer, multi-site scribe trials that measure note accuracy as well as time saved.
• How the FDA handles generative AI devices. The agency gave a breakthrough designation in March 2026 to a patient-facing application from RecovryAI, the Congressional Research Service says. It published a discussion paper in August 2026, MedTech Dive reported.
• Whether small and independent hospitals close the adoption gap.

This article is informational, not medical advice. Patients with questions about how AI is used in their own care should ask their clinician.