Why Smartphone Melanoma Apps Miss Cancers a Dermatologist Would Catch

Smartphone apps that promise to screen your moles for melanoma consistently miss cancers that a trained clinician would flag, and the best available evidence says the gap is large. When Cochrane reviewers pooled data on automated skin-check apps, sensitivity ranged from as low as 7 percent to about 73 percent, meaning some apps missed most of the melanomas they were shown. A separate systematic review in The BMJ reached the same conclusion in blunt terms, finding that current algorithm-based apps cannot be relied on to detect all cases of melanoma. The gap comes down to thin study data, image conditions the phone cannot control, and marketing claims that ran ahead of the science.

What the accuracy reviews actually found

Two systematic reviews anchor this question. The Cochrane review of smartphone applications for triaging suspicious skin lesions identified only two eligible studies covering six apps. Across the automated, AI-based apps, sensitivity spanned roughly 7 to 73 percent and specificity roughly 37 to 94 percent. Those are not narrow ranges. A test that catches 7 percent of cancers in one setting and 73 percent in another is not a stable diagnostic instrument; it is a coin flip whose bias you cannot predict in advance.

The BMJ systematic review of algorithm-based apps looked wider, across nine studies and six apps. The single most-studied app reached about 80 percent sensitivity and 78 percent specificity for malignant or premalignant lesions under favorable study conditions. Even taken at face value, 80 percent sensitivity means one in five cancers slips through. The reviewers went further, warning that these figures were likely optimistic because the underlying studies were small, methodologically weak, and prone to selective recruitment.

Why the reported numbers flatter the apps

The studies behind these apps share a set of biases that all push accuracy upward. Lesions were often selected because a clinician had already decided they were worth biopsying, so the app was tested on a pre-filtered, easier population than the worried person at home. Images were frequently captured by clinicians under good lighting rather than by users with a shaky hand and a bathroom mirror. And a striking share of images were simply unevaluable, with rates reported from a few percent up to nearly half in some datasets. When an app quietly discards the photos it cannot read, its published sensitivity describes only the cases it chose to grade.

What a trained clinician does that an app does not

The contrast is instructive. In the Cochrane data, the one store-and-forward service that routed images to a dermatologist for review reached about 98 percent sensitivity, far above any automated algorithm in the same review. The lesson is not that phones are useless. The diagnostic work is being done by the human, not the software.

A dermatologist evaluating a lesion is not matching a single photo against a library. They integrate history, how fast a spot changed, whether it looks different from a person's other moles, texture and firmness on palpation, and dermoscopic features invisible to a phone camera. They also know when a benign-looking lesion still warrants a biopsy. An automated app compresses all of that into pixels from one uncontrolled image, then returns a risk category with a confidence the underlying evidence does not support.

The regulatory gap the reviewers flagged

Part of why these apps reached consumers with strong promises is regulatory. Under the framework the BMJ reviewers examined, a skin-risk app marketed in Europe could carry a CE mark as a low-risk medical device through self-certification, without independent inspection of its accuracy. The reviewers concluded plainly that this process did not provide adequate protection to the public.

Regulators have acted where marketing crossed into deception. In 2015 the United States Federal Trade Commission brought cases against the marketers of two melanoma-risk apps, charging that they claimed to assess melanoma risk, including in early stages, without competent and reliable scientific evidence. The settlements barred those marketers from claiming any device can detect or diagnose melanoma unless the claim is truthful and backed by human clinical testing of the device, a standard most consumer skin apps have never met.

How to read a skin-check app's claims

Evidence literacy helps here more than any single verdict. A few questions separate a defensible tool from a marketing story. Does the company report sensitivity and specificity from studies where ordinary users, not clinicians, took the photos? Does it disclose how many images were unevaluable and excluded? Is the app cleared or approved as a diagnostic device, or merely self-certified and described with softer words like wellness or education? An app that positions itself as a prompt to see a clinician is being honest. One that implies it can rule out cancer is making a claim the literature does not currently support.

The practical takeaway from both reviews is consistent. These tools may raise awareness or nudge someone toward a visit, but a reassuring app result should never substitute for evaluation of a changing, bleeding, or asymmetric lesion by a qualified professional. This article is educational and is not medical advice.