AI Dermatology Tools Improve Patient Diagnostic Accuracy in Survey Study

Key Takeaways

• AI-assisted dermatology tools were associated with higher patient willingness to name skin conditions and improved diagnostic accuracy vs standard web search, new research suggests.
• Accuracy gains were greater when condition suggestions aligned with dermatologist-provided differentials.
• Improvements in next-step decision-making were modest.

An AI-powered dermatology application was associated with improved consumer understanding of skin conditions, according to new research published in JAMA Dermatology. 

The study included 2,345 US participants who had sought information about a skin concern within the prior year. Participants were randomized to a control group using standard web-based tools, an AI group receiving predictions from a prototype application, and a “Wizard of Oz” group presented with dermatologist-curated differential diagnoses through the same interface. Participants reviewed deidentified images and structured histories acrpss 11,725 case evaluations and reported suspected diagnoses, next steps, confidence, and satisfaction.

Participants in both the AI (62.26%) and Wizard of Oz (61.76%) arms were more likely to name a condition versus controls (41.21%; P < 0.001). Diagnostic accuracy improved from 7.86% in the control group to 22.79% in the AI arm (P < 0.001) and 36.20% in the Wizard of Oz arm (P = 0.002). Improvements in next-step accuracy were more modest, with only the Wizard of Oz arm demonstrating a statistically significant increase vs control (62.95% vs 60.10%; P < 0.001).

The findings suggest that AI-supported tools may enhance patient engagement and understanding, though performance remains closely tied to the accuracy of presented differentials. Notably, even when aligned with dermatologist input, residual inaccuracies highlight limitations in how information is conveyed.

“AI applications were associated with increased accuracy and confidence of consumer understanding of skin concerns, with the degree of improvement in accuracy varying by the accuracy of presented conditions,” the authors wrote. “Benefits further improved when predictions were as accurate as possible.”

Source: Sayres R, Jain A, Venkatraman M, et al. JAMA Dermatol. Published online April 15, 2026. Doi:10.1001/jamadermatol.2026.0597