A recent study in JAMA Dermatology provides a realistic view of AI versus human performance in diagnosing skin cancer.
Key finding: AI surpassed dermatologists with <3 years of experience, matched those with 3–10 years, but remained inferior to experts with >10 years of experience.
The Study
Researchers tested modern AI models (including foundation models) against 652 physicians on 1,117 realistic dermatology cases with clinical and dermoscopic images.
Top AI accuracy: 72%
Humans by experience:
<3 years: ~68%
3–10 years: ~72%
10 years: 74.2% (highest overall)
Why Experience Still Wins
Seasoned dermatologists integrate subtle clinical context, patient history, and rare presentations that current AI struggles with. AI excels as a supportive tool - especially helpful for novices, triage, education, and areas with limited specialist access - but does not yet replace expert judgment.
Takeaways
Patients: Prioritize experienced specialists for concerning skin lesions. AI may improve initial screening and access, but expert care remains essential for early skin cancer detection.
Physicians: Use AI as a collaborator to reduce workload and support trainees, while continuing to build your own expertise.
Future: Hybrid human-AI systems are the path forward.
What are your thoughts - will AI close the gap soon?
References:
Anriot et al. JAMA Dermatology (2026).
https://jamanetwork.com/journals/jamadermatology/fullarticle/2849416