Where do doctors use AI?

The chatbot will see you now.

1. China Leads Clinicians' AI Use

According to Elsevier's Clinician of the Future 2025 survey of 2,206 clinicians across 109 countries, 48% now use AI for work, nearly double the 26% a year earlier.

🇨🇳 China leads the survey at 71%.

At the other end, 🇺🇸 the United States sits at 36% and 🇬🇧 the United Kingdom at 34%, the two lowest rates recorded.

Nearly all clinicians who use AI (97%) have tried a general-purpose tool such as ChatGPT. Looking ahead, 70% expect clinical AI tools to save them time, 58% expect faster diagnoses, and 54% expect more accurate ones.

💡 Worth noting: the 37-point gap between China and the UK is larger than the UK's entire adoption rate.

2. India's Patients Lead the World

Turn from clinicians to patients, and a different country comes out on top. Boston Consulting Group surveyed more than 13,000 adults across 15 countries about using AI for their own health, and 57% said they do.

🇮🇳 India leads at 85%, followed by 🇧🇷 Brazil at 67% and 🇨🇳 China at 62%.

🇺🇸 The United States comes in at 50%, 🇩🇪 Germany at 44%, 🇬🇧 the United Kingdom at 43%, and 🇯🇵 Japan at 34%.

Use falls steeply with age: 78% of Gen Z and 71% of Millennials use AI for health, compared with 50% of Gen X and 31% of Baby Boomers.

Chatbots are the most common entry point, used by 33% of respondents, ahead of AI-enabled wearables at 19%. Some 55% want AI to book appointments and manage referrals, but only 22% are comfortable with AI triaging a heart attack.

🔎 The gap worth sitting with: India's 85% is exactly two and a half times Japan's 34%.

3. What Clinicians Actually Use It For

Philips' Future Health Index 2026 asked more than 2,000 healthcare professionals across 10 countries how they use AI day to day.

Fifty-two percent use it to transcribe clinical notes, 46% use generative AI as a professional "buddy" to talk through ideas, 45% use it to suggest diagnoses from symptoms, and 44% use it to flag dangerous drug combinations.

Nearly six in ten (58%) report workflow efficiency gains, and more than a third say AI now lets them see more patients, by a median of five extra patients a week.

📊 Worth remembering: note-taking leads the list, seven points ahead of diagnosis suggestions.

4. 1,614 Medical Devices and Counting

🇺🇸 The United States has now authorized 1,614 AI-enabled medical devices, up from 1,451 a year earlier, according to The Medical Futurist's analysis of the FDA's latest annual update.

Radiology accounts for 1,230 of them, about 76%, with cardiovascular a distant second at 154 devices and neurology third at 76.

The pace keeps climbing: the FDA authorized 6 AI devices in 2015, 225 in 2023, 235 in 2024, and 335 in 2025.

🔮 A trend worth watching: as of September 2026, not one device on the FDA's list uses generative AI or a large language model.

5. Sweden's Mammogram Trial

🇸🇪 Sweden's MASAI trial randomly assigned more than 100,000 women at four screening sites to either AI-supported mammography or standard double reading by two radiologists.

AI-supported screening detected 29% more cancers and cut radiologists' reading workload by 44%. In the two years afterward, women in the AI group were diagnosed with 12% fewer cancers between screenings, including 27% fewer aggressive subtypes.

The AI triaged low-risk scans to a single radiologist and high-risk scans to two.

🔎 The takeaway worth holding onto: it was the first randomised controlled trial of its kind, and the extra detections came without a rise in false positives.

6. Britain's Own Test

🇬🇧 The United Kingdom ran its own evaluation in NHS Grampian, where an AI tool supported the screening of 10,889 women and raised cancer detection by 10.4%.

The same evaluation, published in Nature Cancer in March 2026, found the tool could cut healthcare workers' workload by more than 30% and shorten the time to notify affected women from 14 days to 3.

Every mammogram in the UK programme is currently read by two radiologists, across more than 2 million examinations a year, and a larger trial called EDITH will test the approach at sites across the country.

🔎 The honest takeaway: with 2 million mammograms a year each read twice, a workload cut of 30% or more would free up a great deal of specialist time.

7. American Physicians Say Yes, and Worry

🇺🇸 In the United States, 81% of physicians now use AI in their practice, according to the American Medical Association's 2026 survey, more than double the 38% recorded in 2023.

Doctors now report an average of 2.3 uses each, up from 1.1 in 2023, led by summarizing medical research and documenting clinical care.

More than three-quarters say AI improves their ability to care for patients, up from 65% in 2023, and 70% see it as a way to automate tasks that contribute to burnout.

Yet 88% are concerned about potential skill loss, particularly among physicians with 10 years or less in practice.

👩‍⚕️ The bigger picture: about 40% of physicians say they feel equally excited and concerned, a sign that fast adoption and open worry are moving together.