Google's research team published a formal study this week in JAMA Dermatology testing a straightforward question: can an AI tool help regular people understand what is happening with their skin? The answer is yes, partly.
The study showed 2,345 people real skin condition cases with photos and asked them to research the condition as if it were their own. Those using the AI tool were willing to attempt a diagnosis at a much higher rate: 62% tried, versus 41% in the group using standard web search. More importantly, the accuracy when they did guess jumped from 8% in the standard search group to 23% with the AI. That is a meaningful lift.
The more interesting result is what the AI did not fix. Deciding what to do next, whether a condition needs urgent care, a routine appointment, or just a pharmacy visit, barely improved. Users with AI assistance were actually slightly more likely than the control group to suggest a less urgent course of action than a specialist would recommend. They felt more confident, but that confidence was not always warranted.
This gap matters. A person who correctly identifies a condition but underestimates how serious it is may delay care that should happen quickly. Confidence without accurate urgency assessment is its own risk.
Google also ran a separate, smaller real-world study with 110 participants who had actual skin concerns. Clinicians said the AI's suggestions matched their own assessments 86% of the time, and found the app useful 92% of the time as a shared reference point during consultations. Patients' ability to name their condition improved by 260% when using the app, though the overall rate of correct guesses was still low in absolute terms.
This research sits inside a real and growing problem. Dermatologist waiting times in many countries run from weeks to months. In one study, 55% of patients waited more than three months for an appointment. Skin cancer rates are rising: new invasive melanoma cases in the US alone rose 42% between 2015 and 2025. Meanwhile, people are already turning to the internet and AI for answers, whether or not those tools are designed for that purpose. Google alone records around ten billion annual searches related to skin, hair, and nails.
The AI skin analysis market reflects this demand. It was valued at around $1.6 billion in 2025 and is projected to reach nearly $8 billion by 2035. The tools exist, the demand exists, and the access gap is real.
For anyone running a business that involves employees, customers, or insurance: this research signals that AI-powered health guidance tools are moving into mainstream use faster than most healthcare systems are ready for. The tools work better than nothing. But they do not yet tell people when something is serious.
The practical implication is simple. If your employees, your customers, or your dependents are using AI to self-triage skin concerns, the tool may help them name the condition and feel reassured. It may not prompt them to act quickly enough when they should. That is the design problem Google is still working on, and it is the right problem to be working on.