AI vs Rheumatologist? The Better Answer May Be Both Save
Can artificial intelligence answer patients’ rheumatology questions as well as a rheumatologist? Abstract 2965 at #ACR26 suggests the answer may be yes, but also highlights why physician oversight still matters.
Investigators evaluated 20 common questions identified by patients with rheumatic diseases. Responses were generated by rheumatologists, three AI platforms, ChatGPT-5.1 Thinking, Gemini 3 Pro, and OpenEvidence, and AI followed by rheumatologist review. Twenty patients and 12 healthcare professionals then rated the blinded responses for measures including readability, utility, reliability, and completeness.
Patients rated all three approaches similarly, with overall scores of 4.2 for rheumatologist-written responses, 4.2 for AI, and 4.3 for rheumatologist-reviewed AI.
Healthcare professionals, however, saw an advantage to the hybrid approach. Rheumatologist-reviewed AI scored significantly higher overall than AI alone and performed better in domains including readability, utility, and completeness. Interestingly, Gemini required the fewest physician modifications, while ChatGPT required the most.
The most important message here is not that AI can replace the rheumatologist. Rather, it may change how we use our time. If AI can rapidly generate a high-quality first draft of patient education while the rheumatologist verifies accuracy, adds clinical context, and personalizes the response, that could make communication both more efficient and scalable.
This was a small study evaluating perceived quality, not clinical safety or patient outcomes. Still, it offers a glimpse of a realistic future for rheumatology: not AI versus physician, but AI plus physician.
More ACR Coverage:
Related Content
EULAR works with the network of national organisations of People with Arthritis/Rheumatism in Europe (PARE) to ensure the voices of people with RMD are heard and have influence – creating powerful alliances that make a real difference. Artificial intelligence is another potential ally in that fight. At the EULAR 2026 Congress in London, work was presented from three PARE initiatives that showcase how digital tools can be used in treatment education and decision-making, and potentially avoid the use of general web searches to answer important questions.
Prof. Valmed, a large language model (LLM) cleared by European regulators for medical diagnosis, outperformed human physicians for diagnosing rheumatologic conditions in terms of speed but not accuracy, a randomized trial showed.
If you are a health practitioner, you may Login/Register to comment.
Due to the nature of these comment forums, only health practitioners are allowed to comment at this time.