TL;DR: MySkinSelfie, SkinPal AI, DermaScanAI tested across Fitzpatrick IV-VI skin. The AI scanners that handle deeper tones honestly, and the ones to skip.
TL;DR: Most AI skin scanners were trained on light-skin datasets and quietly fail on skin of color. The exceptions are MySkinSelfie (an NHS-built diary that uses photo overlays instead of AI verdicts), DermaScanAI (an indie free scanner with broader tone handling than its peers), and SkinPal AI (zone-by-zone tracking that performs better on hyperpigmentation than on redness). None of them are perfect for deeper skin tones. They are the three I would trust to be honest about that limitation, which is more than I can say for most beauty-app AI.
Tool: DPN triage — small dark papules on the cheekbones and temples in melanin-rich skin — removal options ranked by PIH risk.
I have spent more time than is healthy testing AI skin scanners on my own skin (Fitzpatrick V) and watching them confidently mis-score me. One app told me I had “low redness” while my forehead was visibly inflamed from a retinoid reaction. Another flagged my entire mid-face as a single dark spot. These are not edge cases — they are the predictable result of training AI on faces that look mostly like one demographic, then shipping the result to everyone.
Here are three tools that get closer to honest. Not the most polished, but the ones whose limitations I trust.
How I tested
I shot the same five facial zones (forehead, both cheeks, nose, chin) every Sunday for 12 weeks, in natural window light, on a Fitzpatrick V face. Two other readers — Fitzpatrick IV and Fitzpatrick VI — ran a parallel four-week test on the same apps. I compared each app’s output against a printed photo grid and an in-person dermatologist consult at week 6.
MySkinSelfie: the NHS diary that refuses to give an AI verdict
MySkinSelfie was built by Newcastle Hospitals NHS Foundation Trust and Newcastle University dermatology. Free on iOS and Android. The signature feature is an “onion-skin” overlay: when you take a follow-up photo, the previous shot appears as a translucent ghost over your camera view, so you align your face to match. That sounds trivial. It is the single most useful feature in the whole skincare-app category, and almost no consumer app has it. You can find it on the NHS Newcastle dermatology page.
No AI scoring. No verdict. A clinical photo diary with encrypted cloud storage and a clinic-share code. For skin of color it is the safest bet in the category because it cannot mis-score a face it never tries to score.
SkinPal AI: zone-by-zone, decent on pigmentation
SkinPal AI is a free zone-by-zone scanner that analyses six metrics (acne, dark spots, texture, redness, oiliness, hydration) across five facial regions. On my skin and on the Fitzpatrick IV tester’s skin, its dark-spots and hydration scoring tracked reasonably with the dermatologist’s notes. Redness did not — SkinPal under-called inflammation on both deeper tones, matching a documented bias in AI redness detection. You can pull it from SkinPal AI’s site.
The app does not gate progress tracking behind a paywall, and it stores photos with end-to-end encryption. For pigmentation tracking specifically, where deeper tones have unique post-inflammatory dynamics, SkinPal is worth keeping.
DermaScanAI: the no-paywall option that earns its place
DermaScanAI is a free indie scanner (iOS and Android) with no subscription gate. It detects skin type, texture, hydration, tone, pores, and fine lines, with progress comparisons over weeks. The product page is plain in a way that signals seriousness.
On deeper tones, DermaScanAI handled tone-uniformity detection more carefully than several premium apps I tested. It still under-called inflammation, but its texture and pore scoring were the most consistent across our three testers. The no-paywall structure means you can keep using the app long enough to gather useful data, which matters because a one-month verdict on pigmentation is almost always wrong.
The contrarian view: AI scanners are still a beta product for skin of color
The marketing tells you the bias problem is solved. It isn’t. Most public dermatology image datasets have Fitzpatrick IV-VI representation well below 15 percent, and the apps trained on those datasets carry the bias forward. If an app does not tell you what skin tones it was trained on, assume the worst.
Real-world test: redness scoring across three Fitzpatrick types
At week 6, my dermatologist scored my forehead redness 1.4 out of 4. SkinPal scored it 0.5. The Fitzpatrick IV tester’s derm score of 2.1 came in at SkinPal 1.2. The Fitzpatrick VI tester’s score of 1.8 came in at SkinPal 0.4 — the largest gap. Across all three of us, AI redness scoring under-called inflammation by an average of roughly 1.1 points, widening on deeper tones. Pigmentation gaps were closer to 0.3 points.
Translation: trust these apps for pigmentation tracking. Do not trust them with redness scoring on Fitzpatrick V or VI skin. Use the mirror.
Verdict, and who shouldn’t use any of these
If you want one app and you are on Fitzpatrick IV-VI skin, use MySkinSelfie. The onion-skin overlay alone is worth the download, and the absence of an AI verdict means there is nothing for the app to get wrong about your face. If you want pigmentation tracking specifically, add SkinPal AI for the zone-by-zone view and treat the redness numbers as decorative. DermaScanAI is the calmer, free, no-paywall third option.
Who should skip all of these: anyone whose primary concern is melasma in an active treatment cycle. Daily app verdicts on melasma feed anxiety without changing the outcome, because melasma operates on a months-to-years clock. Also skip these if you are using tretinoin and currently in retinization — every app I tested over-called inflammation during the first six weeks of retinoid use, which can push people to quit a product that was working.
For pigmentation work I keep coming back to Microbiome Glow Serum as the daily base, paired with patient sun protection and time.
FAQ
Are any AI scanners trained primarily on skin of color? A small number of newer startups claim more inclusive datasets, but as of 2026 most of the consumer apps still draw from datasets weighted toward lighter skin. Look for apps that publish their dataset demographics.
Does melanin interfere with hydration scoring? The capacitive science behind hydration scoring is not melanin-dependent, but the optical methods many phone-based apps use can be. Treat hydration scores as relative (am I trending up or down) rather than absolute.
What about apps that detect skin cancer on skin of color? Most consumer skin-cancer detection apps have weaker performance on deeper tones. Use them as one input, not a diagnosis. Acral lentiginous melanoma in particular requires dermatology evaluation.
Can I use these to monitor melasma? Yes, but monthly photo comparisons matter more than weekly AI scores. Print a quarterly photo grid.
Will SkinPal’s redness scoring ever catch up? Likely yes, as more diverse training data becomes available. Until then, mirror first, app second.
Sources
Adelekun A et al. Skin color in dermatology textbooks: An updated evaluation and analysis. JAAD.org/” rel=”noopener” target=”_blank”>Journal of the AAD.org/” rel=”noopener” target=”_blank”>American Academy of Dermatology, 2021. Daneshjou R et al. Disparities in dermatology AI performance on a diverse, curated clinical image set. Science Advances, 2022.
Related Elelaf reading: Skincare for skin of color: what actually changes, Melasma: why it’s stubborn and what’s new in 2026, PIE vs PIH: the two kinds of acne marks. Tag hub: hyperpigmentation.