The Latest AI in Healthcare Statistics (2026)

Research AI in Healthcare
By Dr. Jason Lee in New York City
Last updated: September 30, 2026

81% of physicians now use AI professionally. In 2023, it was just 38%.

Hospitals and health systems are moving quickly, too. Half of surveyed U.S. healthcare organizations have implemented generative AI, meaning tools that can create text, images, or other content. And 62.6% of hospitals using Epic, the most widely used electronic health record (EHR) system in U.S. hospitals, have adopted ambient AI documentation tools, which draft clinical notes from conversations between clinicians and patients.

AI can help with documentation, scheduling, billing, research, and patient communication. However, there are real concerns about accuracy, privacy, and trust. And 92% of physicians say they want more training on how to use it.

So is AI actually making healthcare better? Or is it creating new problems? What are hospitals and clinicians using it for, which tools are they choosing, and what do they think of it all?

This article covers 52 statistics, trends, and surveys on the state of AI in U.S. healthcare in 2026.

Editor’s Picks

  • 81% of physicians use AI professionally.
  • 62.6% of U.S. hospitals using Epic have adopted ambient AI documentation tools.
  • 50% of surveyed U.S. healthcare organizations have implemented generative AI.
  • Around one-third of large healthcare organizations are operating AI at scale.
  • 49% of large healthcare organizations are still experimenting with generative or agentic AI.
  • 39% of physicians use AI to summarize medical research or standards.
  • 27% of surveyed physicians reported no AI training from any source.
  • 70% of physicians think AI could automate tasks that contribute to burnout.
  • 19% of surveyed healthcare organizations have implemented agentic AI.
  • 82% of healthcare leaders expect positive ROI from generative AI.
  • 34% of U.S. adults have used an AI chatbot for a health-related purpose.
  • 72% of Americans want healthcare providers to tell them when AI is being used in their care.

Hospital AI Adoption Statistics

71% of surveyed U.S. hospitals used predictive AI integrated with their EHR in 2024 (HHS/ASTP–ONC)

That’s up from 66% in 2023. The federal analysis looked at 2,080 non-federal acute-care hospitals. Hospitals that didn’t answer, or said they didn’t know, were excluded.

Predictive AI uses statistical analysis and machine learning to estimate risks, like which patients are likely to be readmitted or miss an appointment. Note that this figure is about adoption. It doesn’t tell us anything about market revenue or about generative AI.

Hospital and Health System AI Use Statistics

50% of surveyed U.S. healthcare organizations have implemented generative AI (McKinsey)

That’s according to McKinsey’s latest survey of U.S. healthcare leaders, conducted in fall 2025. And most of them have moved past the testing stage. More than 80% of organizations that had implemented generative AI had already rolled out their first use cases to end users.

Around one-third of large healthcare organizations operate AI at scale (Deloitte)

Around one-third of large healthcare organizations operate AI at scale (Deloitte)

In Deloitte’s 2026 outlook survey, published in December 2025, another 49% were still experimenting with generative or agentic AI (AI that can carry out multi-step tasks on its own). The remaining 18% hadn’t adopted either technology.

In other words, most large healthcare organizations are using AI in some form, but only a minority have made it a core part of how they operate.

62.6% of U.S. hospitals using Epic have adopted ambient AI documentation tools (American Journal of Managed Care)

That works out to 1,744 Epic hospitals.

Ambient AI tools listen to conversations between clinicians and patients and turn them into draft clinical notes. As we discuss below, just three products account for most of these implementations.

Nonprofit Epic hospitals are 2.4x more likely to have adopted ambient AI than for-profit hospitals (American Journal of Managed Care)

70.2% of nonprofit Epic hospitals have adopted ambient AI, compared with just 28.8% of for-profit hospitals.

Ownership was one of the biggest differences the researchers found.

64.7% of metropolitan Epic hospitals have adopted ambient AI (American Journal of Managed Care)

That compares with 54.3% of hospitals outside metropolitan areas.

Financial health also played a role. Hospitals with stronger operating margins were more likely to adopt ambient AI.

And adoption reaches 73.1% among Epic hospitals with the highest staffing-adjusted workloads (American Journal of Managed Care)

Among hospitals in the lowest workload quartile, that figure is 61.7%.

Unsurprisingly, hospitals where staff are most stretched appear more likely to turn to ambient documentation tools.

Why Healthcare Organizations Use AI

54% of care-delivery organizations have implemented generative AI for clinical productivity (McKinsey)

Clinical productivity is one of the most common uses of generative AI among care-delivery organizations.

These tools can help with documentation, summarizing information, and other parts of the job that take up a lot of staff time.

44.7% of AI-documentation users reported less EHR frustration in one trial (JAMA Network Open)

That compares with just 14.5% in the control group. In this nonrandomized trial at Atrium Health, 38 of 85 survey respondents using an AI documentation tool reported less frustration with their EHR, versus 8 of 55 who weren’t using one.

The study measured self-reported changes after five weeks. It’s not a national estimate, and it doesn’t prove that every AI tool improves the EHR experience.

Changes in AI Use in Healthcare

81% of physicians use AI professionally, up from 38% in 2023 (AMA)

81% of physicians use AI professionally, up from 38% in 2023 (AMA)

Usage rose to 66% in 2024, before reaching 81% in 2026.

That’s an increase of 43 percentage points in just three years. Put another way, physicians are more than twice as likely to use AI now as they were in 2023.

Physicians now use an average of 2.3 AI use cases (AMA)

In 2023, the average was 1.1.

So physicians aren’t just more likely to use AI. They’re also using it for more things.

Generative AI implementation doubled between 2023 and 2025 (McKinsey)

In late 2023, 25% of surveyed healthcare organizations had implemented generative AI. That rose to 47% in 2024, and 50% in the fall 2025 survey.

Interestingly, most of that jump happened in a single year. Growth has been much slower since.

Nearly 65% of nurses say they are using more AI than one year ago (McKinsey)

More than three-quarters said they use AI in their daily work to some degree.

As we discuss below, though, how much nurses use AI varies widely.

AI Training, Trust and Safeguards

27% of surveyed physicians reported no AI training from any source (AMA)

The other 73% had received at least some training. That’s based on 1,419 physicians who answered the training question in the AMA’s 2026 survey.

Note that this measures whether physicians got any training at all, not whether it was adequate.

92% of physicians want more education and training on AI (AMA)

It’s clear the demand is there. Physicians most often said clinical evidence and practical implementation guides would help them evaluate and use AI tools.

88% of physicians say safety and efficacy validation is important for AI adoption (AMA)

Data privacy matters almost as much, with 86% saying privacy assurances are important.

And 85% want physicians to be consulted, or directly involved, when their organization decides which AI tools to adopt.

72% of Americans want providers to disclose when they use AI (Pew Research Center)

That’s true for both clinical and administrative uses of AI.

81% of Americans want to be told when AI is used to make a diagnosis (Pew Research Center)

The same share want to know when AI is used to analyze medical scans. And 80% want to be told when AI is used to explain lab results.

56% of Americans want disclosure even when AI is used to schedule an appointment (Pew Research Center)

64% want to be told when AI is used to order prescription refills, and 72% want to know when AI takes notes during an appointment.

Most Americans, then, want transparency even for fairly routine, administrative uses of AI.

How Healthcare Professionals Use AI

How healthcare professionals use AI

39% of physicians use AI to summarize medical research or standards (AMA)

That was the most common professional use of AI among physicians.

Physicians have a large volume of research and guidance to keep up with, and AI can help them work through it more quickly.

30% of physicians use AI for discharge instructions, care plans or progress notes (AMA)

Documentation is one of the biggest areas of physician AI use.

These tools can draft notes, summarize information, and produce other written materials used during or after a visit.

28% use AI for billing, chart or visit-note documentation (AMA)

The same share use AI to create chart summaries.

So several of the most common physician uses of AI are about paperwork, rather than diagnosis or treatment.

19% use AI to draft patient portal responses (AMA)

Another 18% use it for translation.

Patient communication is another area where AI can cut down on the amount of writing physicians and staff have to do by hand.

Just 17% of physicians use AI for assistive diagnosis (AMA)

That’s less common than research, documentation, chart summaries, and other workflow tasks.

It also helps explain why so much of healthcare AI adoption so far has centered on productivity and administrative work.

23% of frontline nurses say they don’t use AI (McKinsey)

On the other end of the spectrum, only around 2% said AI is built into everything they do.

AI use among nurses varies a lot, from none at all to heavy daily use.

Around 1 in 10 nurses are heavy AI users (McKinsey)

McKinsey calls these nurses AI “superusers.” They rated their own AI use between 8 and 10 on a 10-point scale.

Which AI Tools and Platforms Are Healthcare Organizations Using?

DAX Copilot, Abridge and ThinkAndor account for more than 80% of ambient AI implementations among Epic hospitals (American Journal of Managed Care)

These three products were the most commonly adopted ambient documentation tools in the study.

That’s a fairly concentrated market. Just a handful of vendors account for most of the ambient AI tools in use at Epic hospitals.

Hospitals source predictive AI from EHR vendors, third parties and their own teams (HHS/ASTP–ONC)

Among surveyed hospitals using predictive AI in 2024, 80% got it from their EHR developer. 52% used third parties, and 50% developed it in-house. (These categories overlap, since many hospitals use more than one source.)

This federal analysis is based on American Hospital Association survey data. It covers predictive AI across all EHR vendors, not generative AI or Epic hospitals alone.

What Healthcare Professionals Think About AI

76% of physicians think AI improves their ability to care for patients (AMA)

That’s up from 65% in 2023. Unsurprisingly, physician confidence in AI has grown alongside their use of it.

40% of physicians are equally excited and concerned about AI (AMA)

Most physicians don’t see AI as all good or all bad.

Many see a chance to cut down on administrative work, while still worrying about safety, privacy, accuracy, and relying on AI too much.

70% of physicians think AI could automate tasks that contribute to burnout (AMA)

Documentation and other admin tasks take up a big part of the working day for many physicians. That helps explain the strong interest in ambient documentation and other productivity tools.

And 70% are very or somewhat concerned about skill loss among medical students and residents.

AI use is rising quickly, but physicians are still worried about becoming too dependent on it.

More than 80% of nurses think AI could improve patient care to some degree (McKinsey)

But just 16% think AI could improve patient care significantly.

Nurses also have concerns about accuracy, trust, reduced human interaction, and privacy, as we discuss below.

Benefits of AI in Healthcare

Physician burnout fell from 51.9% to 38.8% after 30 days of ambient AI scribe use (JAMA Network Open)

The study included 263 physicians and advanced-practice clinicians across six U.S. health systems. Participants also reported a lower documentation burden after using ambient AI.

Documentation is one of the areas where AI tools have shown their clearest benefits so far.

One ambient AI tool cut the time physicians spent on notes by 9.5% (NEJM AI)

The randomized trial included 238 outpatient physicians. Interestingly, the second ambient AI product in the trial didn’t show a statistically significant reduction.

In other words, how much time AI saves can depend a lot on which product is used.

Ambient AI reduced emergency physician documentation time by 1.6 minutes per note (Annals of Emergency Medicine)

That might not sound like much. But small savings per note can add up over a full week of shifts.

70.1% of patients exposed to ambient AI said it was helpful (JAMIA Open)

And 73.6% said they’d prefer ambient AI to be used during future visits. That suggests many patients are comfortable with AI note-taking once they’ve experienced it directly.

That doesn’t mean patients don’t want to be told, though. As we mentioned above, 72% of Americans want disclosure when AI takes notes during an appointment.

82% of healthcare leaders expect positive ROI from generative AI (McKinsey)

But just 45% say their organizations have already quantified a positive return.

So healthcare leaders are much more confident that AI will pay off eventually than they are able to prove today.

Barriers to AI Adoption

43% of healthcare leaders say risk and safety are major barriers to generative AI (McKinsey)

The most commonly cited concerns are inaccuracies or bias, security risks, and regulatory compliance.

Integration problems and a lack of in-house skills also make it harder to scale AI.

33% of nurses say accuracy and trust are their biggest AI concern (McKinsey)

Another 18% chose reduced human interaction, and 16% chose data privacy.

It’s clear that using AI more doesn’t necessarily mean healthcare professionals are completely comfortable with it.

Future of AI in Healthcare

19% of healthcare organizations have implemented agentic AI (McKinsey)

Another 51% were pursuing proofs of concept.

Agentic AI is much earlier in its adoption cycle than generative AI. But a majority of surveyed organizations are at least exploring it.

61% of healthcare technology executives are building, implementing or budgeting for agentic AI (Deloitte)

Agentic AI is designed to complete more complex, multi-step tasks with less human input than traditional AI tools. Deloitte surveyed 100 health system and health plan technology executives in September 2025.

And 85% of those executives plan to increase their investment in agentic AI (Deloitte)

They expect that investment to grow over the next two to three years. Almost all of them (98%) expect agentic AI to deliver at least 10% cost savings over that period.

More than 80% of healthcare executives expect AI to create meaningful value (Deloitte)

More than 80% of healthcare executives expect AI to create meaningful value (Deloitte)

That’s in line with the optimism healthcare leaders show elsewhere. As we mentioned above, 82% of healthcare leaders expect a positive return from generative AI.

That’s according to Deloitte’s 2026 mid-year outlook. Another third are still in pilot or limited use. And roughly one-third have been paused or abandoned altogether.

33% of healthcare organizations pursuing generative AI are taking a “buy” strategy (McKinsey)

A year earlier, it was 19%. More organizations are choosing ready-made AI products rather than building their own.

Miscellaneous AI in Healthcare Statistics

FDA authorized 258 AI-enabled medical devices in 2025 (Stanford AI Index)

The number of AI-enabled medical devices keeps climbing. Many are designed for medical imaging and other specialized clinical uses.

Around 76% of FDA-authorized AI medical devices are in radiology (Stanford HAI)

Radiology is by far the largest category of regulated medical AI. That’s likely because AI has been used in imaging for much longer than newer generative AI tools.

28% have used one to quickly find health information. 25% have used a chatbot to look into symptoms, and 20% to help understand lab results.

22% of Americans use AI chatbots to learn about treatments (Pew Research Center)

The same share use them to learn more about a diagnosis they got from a doctor. And 22% say they use AI because it offers health information at little or no cost.

Only 16% of Americans say a healthcare provider has used AI in their care (Pew Research Center)

Another 46% aren’t sure whether AI has been used or not.

That uncertainty could matter more as AI gets built into EHRs and other systems patients don’t see.

53% of Americans feel they have little or no say over whether providers use AI (Pew Research Center)

63% would like more say over how AI is used in their healthcare. And 72% think it’s very or extremely important for providers to tell them when AI is being used.

Sanor publishes for general information. This publication is not medical, financial, or investment advice.

Privacy Policy Terms of Service Disclaimer Sitemap

SANOR research follows the principles of ISO 20252:2019 and the ICC/ESOMAR International Code on Market, Opinion and Social Research and Data Analytics, and is conducted in line with our Independence Policy. Our published reports follow our Editorial Standards.

SANOR research is funded through advisory engagements and research licensing. We do not accept vendor sponsorship, paid placement, or advertising.

All publications are provided for general information only. They do not constitute medical, financial, or investment advice and do not account for the circumstances of any specific business or individual.