Comprehensive Summary
In their 2025 study, Zhang and Levin investigated the utility of AI in diagnosing dermatologic skin conditions like melanoma and seborrheic keratosis (SK). The researchers tested 6 online Skin Cancer Diagnosis Applications (SCDAs) and 2 AI Chatbots (AICs). They ran these trials twice and employed a grading system to make differentiations between the platforms. The researchers found that, on the whole, the AICs and the SCDAs did better in the diagnosis of melanoma as compared to the diagnosis of SK. Furthermore, they determined that the platforms Skinner and AI Dermatologist were most successful in making these diagnoses. The platform ChatGPT-4 was found to lag in its performance as compared to Copilot in the diagnosis of melanoma and SK by 30%. Another platform named Aysa (an SCDA) did poorly on melanoma identification compared to both AICs studied, Skinner, and AI Dermatologist. Given these findings, the authors made a special call to patients, noting that they should seek out advice from physicians, rather than solely using AI platforms to make diagnoses at home.
Outcomes and Implications
Given the widespread implementation of AI within the field of medicine, this study reinforces the use—but not total reliance—on technology to make clinical decisions. Given the potential for metastasis of pigmented lesions like melanoma, the at-home usage of SCDAs and AICs is suggested. Certainly, at-home usage is preferential to simply leaving these neoplasms unexamined. Nevertheless, these SCDAs and AICs should consider integrating avenues through which to give patients access to clinical services after examining their lesions (i.e., giving patients notifications imploring them to seek out professional medical advice after an initial opinion is given via the SCDA and/or AIC). In addition, cautionary messages to patients are required to prevent them from solely relying on these SCDAs and AICs to prevent inaccurate and/or delayed diagnoses.