A comparative evaluation finds promising concordance but highlights gaps that require
In the February JDD, authors compared two large language models, ChatGPT and DeepSeek, against the 2021 American Academy of Dermatology and National Psoriasis Foundation guideline recommendations for psoriasis treatment. The study used guideline based questions and blinded dermatologist review to assess whether model recommendations aligned with evidence based guidance.
The two models demonstrated similar overall concordance with guidelines, suggesting future potential as decision support tools. Both models performed well on clearly supported recommendations, while performance was more variable for recommendations based on limited evidence. Inter model agreement was moderate, indicating that different models may give different guidance for complex scenarios.
Things to consider: AI can be a useful adjunct for point of care reference, triage, and education, but it is not a substitute for clinical judgment. Models may underperform when guidance is nuanced or evidence is limited, and model output should be verified against current guidelines before informing treatment decisions.
Read the full JDD article to review methods and case examples. Consider piloting model assisted workflows only with clear clinician oversight, build local validation before integration, and contribute to efforts that define safe, evidence based uses of AI in dermatology.
J Drugs Dermatol. 2026;25(2):133-139. doi:10.36849/JDD.9482
Blog write-up assisted by AI
