That mole you’re not sure about — is it nothing, or worth a dermatologist’s time? Skin cancer is highly treatable when caught early, but appointments are scarce and worry is plentiful. A class of AI smartphone apps promises a first-line answer from your own camera, and in 2026 the evidence on how well they actually work is getting sharper.
What the research shows
A 2026 prospective study published in the British Journal of Dermatology evaluated the widely used SkinVision app in 1,458 people who turned up with lesions of concern — a real-world test rather than a lab demo. Earlier validation work had reported strong numbers for its algorithms, with sensitivity around 96% and specificity around 90–95% in clinical settings.
The bigger picture
Zoom out, though, and the picture is more nuanced. A systematic review spanning more than 70,000 test images found pooled sensitivity of about 0.91 but specificity of only 0.64 — meaning these tools are fairly good at not missing real cancers, but prone to flagging plenty of harmless spots too. For a screening aid, erring toward caution is arguably the right failure mode.
How people use them
The workflow is simple: photograph a lesion, get a risk read, and a recommendation to monitor or get it checked. Used well, the app is a nudge toward a doctor’s visit for the spots that warrant one — and a measure of reassurance for the ones that don’t.
The caveat
This is where care is essential. A reassuring result is not a clean bill of health, and these apps can both miss cancers and overcall benign moles. Image quality, skin tone and lesion type all affect accuracy, and equity across diverse skin remains an active research question. No app should delay seeing a clinician about a changing, bleeding or unusual lesion.
Why it matters
The promise of AI skin checks isn’t replacing the dermatologist — it’s getting more of the right people to one, sooner. Handled as a prompt rather than a verdict, a phone-based screen can shorten the path from “I wonder” to “I got it checked.” The science is clear-eyed about the limits, and that honesty is exactly what makes the tool trustworthy as a starting point.
