EADV Review Finds Major Safety and Transparency Gaps in Mobile AI Dermatology Apps

A comprehensive review from the European Academy of Dermatology and Venereology (EADV) Artificial Intelligence (AI) Task Force suggests that although smartphone applications have become increasingly common in dermatology, most available apps lack strong clinical validation, regulatory transparency, and publicly available evidence supporting their safety and performance.¹

Investigators systematically evaluated dermatology-related mobile applications available through the Apple App Store and Google Play Store across 44 European countries. The review identified 1,746 apps during the initial search in 2024, with 420 meeting inclusion criteria after screening. Because of the rapid growth of AI technologies, investigators also performed an updated search in 2026 focused specifically on AI-based skin cancer screening applications.

Target Audiences and Primary App Functions

The analysis found that dermatology apps are primarily designed for patients rather than health care professionals. Sixty percent of included apps targeted laypersons, while 40% were intended for clinician use. Most applications (94%) were free to download, and 61% were available on both major app platforms. Teledermatology represented the largest category of apps (54), followed by general dermatology education (51), disease-specific guides (47), and calculators or measurement tools (37). Self-diagnosis applications accounted for 33 apps, with AI incorporated into the majority of those tools.

Overall, only 12% of the reviewed apps reported using AI. Among AI-enabled apps identified in the 2024 dataset, approximately 41% focused on skin cancer diagnosis. The updated 2026 search identified 38 AI-based skin cancer screening applications currently available in Europe.

Gaps in Clinical Evidence and Regulatory Compliance

Despite the expanding number of digital tools, investigators found relatively little evidence supporting their clinical use. Only 20% of all apps were linked to at least 1 peer-reviewed publication indexed in PubMed. However, much of this literature consisted only of brief mentions rather than formal evaluation. Twenty-four apps had been studied in clinical trials, only 5 had undergone randomized controlled trials, and just 6 had been evaluated in real-world post-deployment studies.

Regulatory transparency was similarly limited. Only 2% of all included apps reported having a CE (Conformité Européenne) marking, indicating compliance with European medical device regulations, while only 10% stated that they complied with the General Data Protection Regulation (GDPR).² Among AI-based skin cancer screening applications, relatively few reported appropriate regulatory certification despite functioning as diagnostic support tools. The investigators also evaluated AI skin cancer applications against recommendations developed by the EADV AI Task Force. Compliance with many recommended best practices was inconsistent or poorly documented.

Key Challenges

The strongest areas of adherence involved cost transparency and privacy. Most AI apps disclosed download costs (92%), and 87% provided information regarding user privacy. However, substantially fewer documented key elements related to responsible AI development and deployment.

• User consent for commercial data use: Described by 37%
• Clinical validation disclosure: Disclosed by 32%
• Diverse skin tone representation in datasets: Reported by 21%
• Comparable performance across populations/devices: Addressed by 13%
• Post-market monitoring: Described by only 8%

Other recommended practices, including algorithm explainability, accessible design, avoidance of excluding patient subgroups, and mechanisms for users to report errors, were addressed by fewer than 1 in 5 applications. Multidisciplinary involvement, including participation by dermatologists during development, was reported for 42% of AI apps.

The authors note that limited public information made evaluation challenging. Most developers did not respond to requests for additional documentation, requiring investigators to rely primarily on app store descriptions and publicly available websites when assessing compliance. The review also highlights concerns about representation of diverse skin types in AI training datasets. Previous studies have shown that many dermatologic AI algorithms are developed using predominantly lighter skin tones, raising concerns about reduced diagnostic performance in patients with darker skin. The current review found relatively little evidence that developers had adequately addressed this issue.

Recommendations for Safer Digital Care

Although the authors acknowledge that dermatology apps have significant potential to improve education, teledermatology, and patient engagement between office visits, they conclude that stronger quality assurance is needed before these tools can be relied upon more broadly. They recommend greater adherence to EADV AI Task Force guidance, increased clinical and real-world validation, improved transparency regarding AI development and performance, broader inclusion of diverse patient populations, and stronger regulatory compliance to improve both patient safety and clinician confidence.

References

1. Sangers T, Shifai AN, Pasqualini DGJ, et al. Smartphone Applications in Dermatology With Special Focus on AI: A European-Continent-Based Review. Int J Dermatol. Published online July 15, 2026. doi:10.1111/ijd.70582

2. European Commission Medical Devices EUDAMED. Overview European Commission Health and Food Safety. (2025). https://health.ec.europa.eu/medical-devices-eudamed/overview_en