Skin cancer detection apps exist, and some perform reasonably well in studies, but none of the consumer apps available on your phone have FDA approval for diagnosing skin cancer. Their accuracy varies wildly depending on the app, your skin tone, the type of lesion, and even the quality of the photo you take. Before you trust one with a potentially life-changing decision, here’s what the evidence actually shows.
How These Apps Work
Most skin cancer apps ask you to photograph a mole or spot, then use machine learning algorithms to analyze the image for patterns associated with melanoma or other skin cancers. The app compares your photo against a database of thousands of skin lesion images and returns a risk assessment, usually something like “low risk,” “medium risk,” or “high risk.” Some apps use color and pattern recognition, while others incorporate additional clinical factors like your age or the lesion’s growth history.
This is fundamentally different from what happens in a dermatologist’s office. A doctor can touch a lesion, examine it under a dermatoscope (a magnifying tool with polarized light), and factor in details like texture, scaliness, and your full medical history. Apps working from a single smartphone photo simply can’t capture all of that information.
Accuracy Numbers From Clinical Studies
The best-performing apps catch most skin cancers in controlled research settings, but they also flag a lot of harmless spots as suspicious. One well-studied algorithm achieved 95.1% sensitivity (meaning it correctly identified 95 out of 100 skin cancers) with a specificity of 78.3% (meaning about 22 out of 100 benign spots were incorrectly flagged). Another app, Moleanalyzer pro, hit 95% sensitivity and 76.7% specificity in one trial, and performed comparably to or better than dermatologists in some studies.
Those numbers sound encouraging, but context matters. These results come from carefully controlled studies using curated image sets, not from real people snapping photos in their bathroom with uneven lighting. And when one study tested apps broadly, they missed 41% of melanomas. That’s a staggering failure rate for the deadliest form of skin cancer.
For comparison, dermatologists in clinical studies achieve around 87.5% sensitivity and 81.4% specificity. Primary care doctors land around 79.9% sensitivity and 70.9% specificity. The best apps in ideal conditions approach dermatologist-level performance on paper, but real-world use introduces problems that research settings eliminate.
Where These Apps Fail
Researchers at the University of Birmingham identified three major failings in skin cancer apps: a lack of rigorous published trials proving they work, a lack of specialist input during development, and flaws in the photo analysis technology itself.
The technology struggles most with lesions that don’t look like “typical” skin cancer. Scaly, crusted, or ulcerated spots give the algorithms trouble. So do amelanotic melanomas, which are melanomas that don’t produce the dark pigment most people associate with skin cancer. These unpigmented melanomas are already easy to miss with the naked eye, and apps are particularly bad at catching them.
Without dermatologist input during development, apps may also miss rarer or unusual cancer presentations. And there are clinical red flags that no photograph can capture. If you’re over 40 and develop a new mole that’s been growing, that context alone is medically significant, but an app analyzing a single image has no way to weigh it properly.
Photo quality introduces another layer of unreliability. Lighting, camera angle, skin glare, and even pen marks on the skin all affect results. In one study, marking the skin around a lesion dropped an app’s specificity from 84.1% to just 45.8%, meaning it started flagging nearly everything as suspicious.
Skin Tone and Representation Gaps
Most of these algorithms were trained primarily on images of lighter skin. The American Academy of Dermatology has called for apps to undergo testing that proves accuracy “in people of all skin tones” before release. Until that happens, people with darker skin have even less reason to trust the results. A spot that looks one way on light skin may present very differently on dark skin, and if the training data doesn’t reflect that, the algorithm’s accuracy drops in ways that aren’t captured by the headline statistics.
The Regulatory Loophole
The FDA requires any device that intends to diagnose or treat a medical condition to receive clearance. But app developers routinely sidestep this by classifying their product as “for medical information” or “for entertainment purposes.” With that classification, the developer doesn’t need to prove the app is accurate or safe. The AAD has pointed out that none of the consumer smartphone apps have received FDA approval, so there’s no independent verification of their claims.
The first AI-powered diagnostic tool the FDA actually cleared is DermaSensor, a handheld device that uses spectroscopy (not a phone camera) to examine lesions at a cellular level. It’s designed for use by healthcare providers, not consumers. It received breakthrough device designation in 2021 and subsequent clearance, but it’s a clinical tool, not something you download from an app store.
What Dermatologists Recommend Instead
The AAD’s position is blunt: giving people unregulated diagnostic technology and letting them decide which result is accurate “can do more harm than good.” A false negative, where the app tells you a melanoma is nothing to worry about, could delay treatment for a cancer where early detection is the difference between a simple excision and a life-threatening diagnosis.
That said, dermatologists aren’t opposed to all skin-related apps. Apps that remind you to check your skin regularly, track changes in moles over time with photos, or remind you to reapply sunscreen get strong endorsement. The distinction is between apps that help you monitor and apps that claim to diagnose. Monitoring tools encourage you to notice changes and bring them to a professional. Diagnostic tools encourage you to skip that step.
A Practical Way to Use Them
If you’ve already downloaded one of these apps, the most reasonable approach is to treat it as a prompt, not a verdict. A “high risk” result is worth acting on quickly. But a “low risk” result should never override your own concern about a spot that’s changed color, grown, started bleeding, or just looks different from your other moles. The ABCDE rule (asymmetry, border irregularity, color variation, diameter over 6mm, evolving appearance) remains a more reliable self-screening framework than any current app.
Some NHS practices in the UK are piloting apps like Map My Mole, where a patient photographs a lesion and a dermatologist reviews it remotely, with results returned in days. This hybrid model, where technology handles the image capture but a human specialist makes the call, is closer to what the evidence supports. The camera is useful. The algorithm making the final decision, without a doctor in the loop, is where the risk lives.