How accurate are AI acne scanner apps?

AI acne apps can count and sort the spots that show in a clear photo, but their accuracy is not well tested, and a photo cannot show how deep a spot goes or what causes it. A 2025 review of research models found that AI judged the severity of skin disease correctly 81 percent of the time on average, and its authors called for better studies in real clinics. The FDA's list of authorized AI medical devices, dated 4 September 2026, has no device with acne in its name. How far you can trust an app depends on what it was built to measure, how good your photo is, and whether the bumps on your skin are acne at all. This guide is one part of the complete acne skincare routine.

What a face photo can show an AI tool

A clear, well lit face photo holds real information about acne. It shows how many spots are on the parts of the face the camera can see, and where they sit: forehead, nose, cheeks, chin or jaw. Blackheads are tiny dark dots. Whiteheads are small white or skin colored bumps. Red bumps with no top and pimples with a white or yellow top each look different again. A photo can also show flat red or brown marks, and dents left by old pimples.

An AI model is a computer program that learned to connect patterns in images with labels, after it was shown a large set of example photos. When it gets your photo, it returns its best match: a count, a type, or a severity level.

A 2025 systematic review (a study that gathers and compares the earlier studies on one question) in the British Journal of Dermatology included 45 studies of AI models that graded how severe a skin disease was from images, and combined the results of 19 of them. The authors' plain language summary says that on average, the AI judged severity correctly 81 percent of the time. The result changed with the disease and with the grading scale. For acne, one accuracy measure, which the authors call sensitivity, was 89.7 percent with one grading scale and 69.8 percent with another. The authors called for high-quality studies in real clinics to produce reliable evidence.

Those numbers describe research models, each tested by its own study team. An app you download may use a different model, tested in a different way or never tested in public.

Photos are useful for showing change over time. When you take the same photo the same way each week, a tool can compare one count with the next. If a tool makes the same error each time, the trend between photos can still be useful even when a single count is off.

What a photo cannot show

The American Academy of Dermatology (AAD) says effective acne treatment "begins with an accurate diagnosis." A dermatologist, a doctor who specializes in skin, diagnoses acne by looking at the breakouts, and also checks whether the bumps are acne in the first place. The AAD names conditions that people mistake for acne: a skin infection called folliculitis, perioral dermatitis, which causes acne-like bumps around the mouth, and hidradenitis suppurativa, which needs different treatment. Rosacea is another. The NHS says rosacea can look a lot like acne, and that it can cause small pink or red bumps, sometimes filled with a yellowish liquid.

A tool built to sort pimples into acne categories has no category for these conditions, so it describes whatever it sees in acne terms. A rosacea bump with a yellow top can be counted as a pimple with pus.

Other facts stay hidden in a photo:

• Depth. A photo shows the surface of the skin. It can show a large, shiny lump, and it cannot measure how deep the lump goes or how much it hurts.
• Hormones. The NHS lists oily skin and acne among the symptoms of polyendocrine metabolic ovarian syndrome (PMOS), a hormone condition, along with irregular periods and extra hair growth. A doctor checks for it with blood tests of hormone levels, and a face photo shows neither a menstrual cycle nor a blood result. See hormonal acne vs regular acne for the pattern.
• Your history. The AAD's list of reasons to see a dermatologist includes breakouts that began after starting a new medicine in the last 6 months, pimple-like bumps in the armpit or groin, and first blemishes at age 8 to 11 or earlier. A face photo shows none of these.

Why light and camera change the result

An AI model has nothing to work with except the pixels in the photo. When the light changes, the pixels change.

The AAD's guide to taking skin photos for a dermatologist explains why light matters. Good lighting "lets your dermatologist see the exact color and texture of your skin," and shadows, glare and harsh lighting "can hide the details." The guide asks for natural, bright light from a window, light that shines on the skin from the front instead of from behind you, all camera filters turned off, and a sharp photo with nothing blurry. A computer model reading the same photo faces the same problem: a spot hidden by glare is missing from its input too.

A 2020 systematic review in The BMJ of phone apps that rate the skin cancer risk of moles found nine small studies of poor quality. In those studies, clinicians chose the spots and took the photos, and many photos could not be assessed at all. The authors wrote that results were likely to be worse when the people who use the apps take the photos. Who took the test photos is a fair question to ask about any photo tool.

How regulators treat skin apps

In the United States, the FDA regulates medical devices, and the legal definition of a device covers products intended for use in diagnosing a disease. The FDA publishes a list of AI-enabled medical devices it has authorized for sale. It says the devices on that list met its premarket requirements, "including a focused review of the device's overall safety and effectiveness."

Software sold for general wellness follows a different path. The FDA's general wellness guidance, issued in January 2026, says software meant to support a healthy lifestyle, and unrelated to diagnosing, treating or preventing a disease, falls outside the device definition. The same guidance says the FDA's device center does not intend to examine low risk general wellness products against the device rules. The intended use, shown by what the maker claims, decides which rules apply.

The version of the FDA's AI device list dated 4 September 2026 held 1,614 devices, and 1,230 of them were in radiology. None had the word acne in its name. The FDA says the list is not complete, because it finds devices mainly through AI terms in their approval paperwork.

Why a dermatologist makes the diagnosis

An AI tool gives a reading of one photo. A dermatologist examines the skin in person, asks about your history and medicines, and can order tests. The AAD says a dermatologist sees which types of acne you have and where they appear, because different types often need different treatment, and that "whether you have acne or another condition, your dermatologist can give you an accurate diagnosis."

UK guidance points the same way. NICE guideline NG198, which tells doctors in England how to manage acne, says to refer a person to a dermatology specialist team when "there is diagnostic uncertainty about their acne," or when they have nodulo-cystic acne, the form with deep lumps and cysts. The NHS tells people to see a GP, a family doctor, if they have tried treating acne themselves and it is not getting better, if they have lots of spots, if spots are deep, painful or causing scars, or if acne is affecting their mental health.

Face photos and privacy

A face photo identifies you. Before you upload one, look for four answers in the app's privacy policy:

• Where the photo is stored, and whether that storage is private.
• Who at the company can open it.
• Whether it is used to train an AI model.
• Whether deleting your account also deletes the photo files.

If the policy leaves any of these unanswered, ask the company before you upload. The AAD makes a similar point about photos you send to a dermatologist: they contain medical information, so ask the office how it keeps them secure.

How to use a photo tool well

A single reading from any tool is one data point. Use the tool for trends:

1. Take the photo in the same place, facing the same window, at the same time of day.
2. Hold the phone at the same distance and angle, or use a small stand.
3. Remove makeup and turn off filters and beauty modes.
4. Compare readings week to week, and look at the photos side by side as well as the numbers.
5. Give a routine time to work. The American Academy of Dermatology says to use an acne product for 6 to 8 weeks before judging it, and that complete clearing generally takes 3 to 4 months.

The full setup is in how to take acne progress photos. Bring your photos to a doctor's visit too. The AAD says good skin photos help a dermatologist tell how well a treatment is working.

What Skinli's photo review does, and its limits

Skinli, which publishes this guide, has its own photo review, so here are its facts in one place. It sends one face photo to a general AI model run by an outside company, with written instructions to look only at the skin, to say nothing about age, race, identity or looks, and to give no diagnosis. The model picks from fixed answers: the main pimple type, a count range (under 10, 10 to 20, or more than 20), marks, where the spots are, and three warning signs (large sore lumps, 10 or more pimples with pus, scars). More than 20 spots, dents or any warning sign puts a "See a doctor first" section right after the score.

Skinli has not measured how its photo review compares with a dermatologist's count, so it publishes no accuracy figure, and it cannot tell acne apart from rosacea or folliculitis. No doctor reviews the photos. The review itself does not save the photo. Saved photos are shown only to you, or to the site owner, after sign-in. Skinli's comparison of AI acne apps sets these limits next to other apps, and the compare page covers prescription services too.