Analysis: ChatGPT and DeepSeek Largely Align With Psoriasis Treatment Guidelines
Key Takeaways
ChatGPT and DeepSeek each demonstrated an overall 87.1% concordance with AAD-NPF psoriasis guidelines, according to a new study.
Performance was strongest for guideline recommendations supported by high-quality evidence.
Chat GPT showed higher alignment with moderate-strength recommentations, and DeepSeek aligned better with limited-evidence recommendations.
Large language models (LLMs) ChatGPT and DeepSeek generally adhere to established psoriasis treatment guidelines, but some performance gaps remain.
Investigators publishing in the Journal of Drugs in Dermatology explored whether two AI models aligned with recommendations from the 2021 American Academy of Dermatology–National Psoriasis Foundation (AAD-NPF) guidelines. To do this, the research team developed thirty-one guideline-based clinical questions reflecting real-world decision-making in psoriasis management. Each question was independently posed to ChatGPT-4 and DeepSeek, and responses were evaluated by two blinded board-certified dermatologists. Answers were classified as concordant or discordant with guideline recommendations, with disagreements adjudicated by a third reviewer.
The results suggested overall concordance with the AAD-NPF guidelines was identical for both models at 87.1%. Performance was strongest for recommendations supported by high-quality evidence, with both ChatGPT and DeepSeek fully aligning with all strong recommendations. Some differences surfaced when guidance was less definitive, however. ChatGPT demonstrated higher concordance for moderate-strength recommendations compared with DeepSeek (87.5% vs 81.3%), while DeepSeek showed better alignment in areas supported by limited evidence (66.7% vs 33.3%). Intermodel agreement was moderate.
"Although ChatGPT and DeepSeek show promise in aligning with evidence-based dermatologic care, inconsistencies, particularly in cases with limited evidence, highlight the necessity of clinical oversight," the authors wrote in the study. "AI may serve as a valuable adjunct in psoriasis management; however, its safe integration into practice requires careful validation and context-sensitive application."
Source: Nebehat A, et al. Journal of Drugs in Dermatology. 2026:25(2)133.