Google Says AI Tripled Skin-Condition Guess Accuracy in 2,345-Person Study
Updated
Updated · Google Research · Jun 12
Google Says AI Tripled Skin-Condition Guess Accuracy in 2,345-Person Study
3 articles · Updated · Google Research · Jun 12
Summary
2,345 survey participants using Google’s dermatology AI were nearly three times as accurate at naming skin conditions as those using standard search tools—23% versus 8%—and were more willing to make a guess.
The JAMA Dermatology study found the tool improved condition recognition, confidence and satisfaction, but not the harder task of choosing what to do next; next-step accuracy in the standard AI arm was not significantly better than control.
Google said that gap reflects the tool’s design: it surfaced possible matching conditions and general dermatologist-written information rather than tailored medical advice, and AI users were slightly more likely to choose a less urgent next step than dermatologists would.
In a separate real-world study of 110 people with active skin concerns, participants’ ability to name their condition rose 260%, while clinicians said the app’s predictions aligned with their assessments 86% of the time and were helpful in 92% of consultations.
Google framed the findings as evidence that image-based, human-centered AI can lower barriers to understanding skin issues, but still needs more personalized and actionable guidance before it can reliably support care decisions.
Google's skin AI is right less than 1 in 4 times. Is this tool empowering patients or creating a new health risk?
If a free AI health tool leads to a delayed diagnosis, who is legally responsible for the resulting harm?
Google's 2026 AI Skin Study: Progress, Pitfalls, and the Path to Equitable Dermatology
Overview
Google's June 2026 study highlights that while AI tools can identify skin conditions, users still struggle to decide on the right medical steps afterward. The study found that standard AI did not significantly improve users' ability to choose next actions, showing that identification alone is not enough. This points to a major gap: people need clearer, more practical guidance from AI to safely navigate what to do next. The report emphasizes that future AI development in skin health must go beyond just diagnosis and focus on supporting users with actionable, trustworthy advice for their next steps.