Physician Use of Augmented Intelligence has Doubled Save

A 2026 Physician Survey to assess artificial intelligence and its use in Medical Practice has been published and follows earlier surveys around adoption and use. This is the third in a series of longitudinal surveys (prior waves 2023, 2024), enabling meaningful trend analysis. AI adoption has more than doubled in three years and physician sentiment shifting decisively toward cautious optimism.

Key Takeaways

• AI adoption among U.S. physicians has doubled since 2023 (81% now vs. 48%). Rheumatologists are not early adopters or nor laggards
• Research summarization AI (used by 39% of physicians, up 26 points since 2024) is the fastest-growing use case — applies to newer therapies and treatment guidelines in rheumatology.
• A minority (17%) us AI to assist diagnosis
• 70–73% of physicians expect AI to automate clinical and administrative tasks linked to burnout
• Skill erosion in trainees is the most widely shared concern (70% very/somewhat worried).
• (Rheumatology fellowship programs should proactively design AI literacy curricula that build, rather than replace clinical reasoning and examination skills.
• Before adopting any AI tool, rheumatologists should demand: (1) peer-reviewed safety/efficacy validation, (2) clear liability frameworks, (3) institutional data privacy assurances, and (4) seamless EHR integration. Vendor trust alone is insufficient.

Summary 

81% of physicians now report awareness or use of AI in professional practice, up from 48% in 2023. Average AI use cases per physician rose from 1.1 to 2.3 over the same period.

The dominant use is heavily documentation-focused, which is directly applicable to rheumatology's notorious documentation load:

• Summaries of medical research and standards of care: used by 39% of physicians — the fastest-growing category, up 26 percentage points since 2024.
• Discharge instructions, care plans, and progress notes: 30% currently; 57–58% enthusiastic and planning near-term adoption.
• Chart summaries and billing code documentation: 28% current use, with rapid growth anticipated.
• Assistive diagnosis: 17% currently, with broad near-term adoption planned, of particular interest in rheumatology, where diagnostic delays of 2–8 years remain common in conditions like axSpA and myositis.

Rheumatology context: The specialties with the heaviest prior authorization and documentation burdens stand to benefit most from AI-assisted clinical documentation. Medical specialties (which include rheumatology) showed personal skill-loss concern at 25%, lower than primary care (34%), suggesting relative comfort with AI integration.

Physicians anticipate net positive impact across nearly every measured domain: work efficiency (78% helpful), diagnostic ability (74%), cognitive load reduction (64%, up from 48% in 2024), and burnout mitigation. The one net-negative domain is patient privacy, the only area where more physicians expect harm than benefit.

Skill erosion is the other major concern: 88% express some level of worry about skill loss, concentrated most heavily on medical trainees and residents (70% very/somewhat concerned). Early-career physicians (<10 years) report significantly higher personal skill-loss anxiety (35%) than late-career peers (21%), a generational divide with implications for rheumatology fellowship training programs.

Physicians generally support patient use of AI for general health and medication questions. But most are wary of patient use for tasks requiring clinical judgment. Nearly half strongly oppose patients using AI to interpret radiology or pathology results.

The survey identifies a clear prerequisite hierarchy for AI trust and adoption. All but two of 21 tested factors were rated important by a majority of respondents:

• Safety and efficacy validation by a trusted entity: 88% rate as important — the top requirement.
• Data privacy assurances: 86% important; concern is significantly higher for non-institutional tools (71%) versus institution-provided tools (42%).
• Clear liability frameworks for AI errors: ranked the #1 regulatory priority by 31% of respondents — ahead of FDA oversight, post-market surveillance, and patient consent requirements.
• EHR integration: seamless embedding in existing workflows is essential, not optional.

On training: 92% want more AI education, yet 27% have received none. Only 11% of those with any training report receiving 'a lot.' Preferred formats are workflow-embedded (alongside EHR training, during practice) and CME-style modules, not training delivered by AI companies themselves, which ranked last.

Physicians also want agency: 85% want to be consulted or responsible for AI adoption decisions in their practice. Peer-reviewed literature and U.S. clinical trial data are the gold-standard evidence formats, not vendor demonstrations.