AI is great at hair care. It is not your dermatologist.

We get this question a lot: "Can your scan diagnose my hair loss?" The short answer is no, and we want to be the loudest voice in the room about why. An AI that looks at a photo of your hair can do some genuinely useful things. Telling you what is wrong with your scalp is not one of them.

What the scan is good at

The Cosmixxology scan takes a guided close-up of your hair and estimates a handful of physical properties: texture and curl pattern, apparent porosity, surface moisture and shine, density, and colour tone. It then matches that profile to ingredients and routines that have worked for similar profiles, and tracks how the same measurements change scan to scan.

These are cosmetic questions. They decide whether your formula leans on aloe and panthenol or on shea and protein, whether a sealing oil goes in, and how much. They are also the questions a photo can reasonably answer: porosity shows up in how the cuticle reflects light, density in how much scalp is visible between strands, and change over time in a sequence of images taken the same way.

That last point is the one we are proudest of. A single scan is an estimate. Six scans over six months are a trend, and a trend is what tells you whether a routine is working.

What it is not

• Not a medical device. The U.S. Food and Drug Administration has a long-standing policy for "general wellness" products: low-risk tools that help you maintain a healthy lifestyle and make no claims about diagnosing, treating or preventing a disease are not regulated as medical devices. Our scan is built to stay inside that line. It has no FDA clearance because it does not claim to do the things that would need one.
• Not a substitute for a dermatologist or trichologist. A clinician can examine your scalp under magnification, do a pull test, take a history, order blood work and, when needed, a biopsy. A phone photo cannot.
• Not reliable for disease. Alopecia areata, telogen effluvium, scarring alopecias, fungal infections, psoriasis and seborrhoeic dermatitis can look alike from the outside and need very different treatment. The scan is not trained to tell them apart and will not try.

Where image AI is known to struggle

This is not just caution on our part. The published evidence on AI in dermatology is clear about where image models fail. In 2022, a Stanford-led team published a study in Science Advances that tested state-of-the-art dermatology AI models on a curated, biopsy-confirmed image set with a full range of skin tones. The models' accuracy, measured as area under the ROC curve, fell by roughly 27 to 36 percent compared with their original test results, with the worst performance on dark skin tones and uncommon diseases. Notably, the dermatologists who label such datasets also did worse on those images.

Hair is not skin, but the lesson transfers directly: a model is only as good as the images it learned from, and the rare and the under-represented are exactly where it is least trustworthy. That shapes how we build. We collect scans across the full range of hair textures and skin tones, we test accuracy by group rather than on the average, and we would rather the scan say "uncertain" than guess confidently on hair it has seen too little of.

When to see a clinician

The American Academy of Dermatology draws a useful distinction between hair shedding, which is usually temporary and often follows illness, childbirth, major weight loss or stress, and hair loss, where something is stopping regrowth and will keep doing so until it is treated. A dermatologist can tell you which one you have, and the Academy's advice is that starting treatment sooner gives better results. Book a real appointment, not just a scan, if any of these is true:

• Sudden, patchy, or rapidly accelerating hair loss;
• Persistent itching, burning or scaling that does not settle with a gentle, fragrance-free routine;
• Visible inflammation, pustules, bleeding or scalp pain;
• Hair changes that come with other symptoms: fatigue, weight change, missed periods, a new medication;
• Any concern that has lasted more than a few weeks. Time matters with scarring alopecias, where follicles that are lost do not come back.

Our scalp comfort guide covers the cosmetic side of itching and flaking, and says plainly where cosmetics stop.

What we do when we are unsure

Every analysis returns a confidence score. When confidence is below our threshold, because of bad lighting, motion blur, an unusual presentation or simply hair the model has seen too little of, we tell you the result is uncertain rather than guessing. You can also request a human review of any analysis from your account, and the reviewer sees the same image and the same scores you do.

If a scan shows a sharp change between sessions, more visible scalp, a texture shift, scores that jump, the app says so and suggests a clinician rather than a new formula. We think that is the most useful thing an AI can do with a problem it cannot solve: notice it early and point you to someone who can.

Educational only, not medical advice. Read our AI transparency disclosure for how the analysis works and what data it uses.