Acne AI still performs worse on darker skin tones, so clinician review remains the safety check for teledermatology patients. Teledermatology means a clinician evaluates acne from photos or video rather than in person.
Detection looks strong in research, but severity grading is weaker. That gap affects treatment choices, referrals, and follow-up timing. A dermatologist can connect the photo score to symptoms and history.
Medical information disclaimer: This article is for general educational purposes only and does not provide medical advice, diagnosis, or treatment. Always consult a physician or other qualified health professional about symptoms, medications, tests, or treatment decisions.
Table of Contents
- Does acne AI detect or grade better?
- Why do darker skin tones get worse results?
- What happens in photo-based care?
- How can you get a safer result?
Does acne AI detect or grade better?
A 2024 systematic review in npj Digital Medicine found deep-learning models diagnosed acne with 94% accuracy but graded severity at only 67-86% accuracy, as detailed in the systematic review in npj Digital Medicine. Triage, the sorting of urgent versus routine cases, is less reliable than detection. A high detection score does not mean correct staging. A 2022 PMC smartphone-acne study reported about 85-91.6% severity accuracy.
It advised teledermatologist referral for high-grade acne rather than standalone diagnosis, especially for darker skin and inflammatory cases. For readers, the takeaway is practical. Use an app score as a starting signal. Let a clinician confirm grade and plan.
Why do darker skin tones get worse results?
A September 2025 PMC/NIH acne severity study found acne models built mainly on Caucasian-patient datasets show significant bias and marked decline on skin of color, as reported in the September 2025 PMC/NIH acne severity study. Training data shapes performance. Underrepresented tones get more errors.
The pattern appears beyond acne. The Stanford DDI team reported state-of-the-art dermatology AI lost 29-40% on ROC-AUC, a standard accuracy score, on pathologically confirmed Diverse Dermatology Images, with dark tones driving the drops. A 2025 evaluation reported in UCLA Proceedings/Cureus found ChatGPT diagnostic accuracy fell 20% for individuals with skin of color.
What happens in photo-based care?
A large teledermatology experiment reported by Northwestern Kellogg in February 2024 found board-certified dermatologists reached 38% top-3 accuracy, meaning the right diagnosis was in their top three guesses, and PCPs reached 19%, as described in the Northwestern experiment report. Both groups did worse on dark-skinned images, though AI support improved accuracy. Photos alone challenge even trained eyes. Regulation still expects human oversight.
FDA records show only one dermatology AI device, DermaSensor for skin-cancer triage in adults 40+, received clearance in January 2024 to assist referrals, and no consumer acne mobile app has FDA approval, according to the FDA clearance reporting via STAT. An app result is not a cleared diagnosis. January 2025 FDA draft guidance directs makers to document training-data demographics, disclose biases, and report subgroup performance. Decision support that a clinician can independently review falls outside device oversight. Ask who reviews your images and how bias was tested.
How can you get a safer result?
Treat the app as an organizer, not a decider. Good photos plus clear history help a teledermatologist correct AI errors.
Share change over time when possible. Request an explicit grade, referral threshold, and follow-up plan in writing. Send new daylight views promptly if acne spreads, darkens, or forms painful nodules.
- Take front, left, and right views in steady daylight, no flash or filter.
- Show one close-up plus one wider view to show spread.
- List duration, pain, new products, medicines, and prior treatments.
- Flag dark spots, deep lumps, scarring risk, or fast worsening.
You Might Also Like
- Skin-Tone Equity Gaps in 2026 Acne AI Research: Privacy and Photo-Data Questions for Teledermatology Patients
- Pediatric Acne and Rosacea at the 2026 AAD Annual Meeting: Evidence and Practice Questions Still Open for Pediatric Dermatology Patients
- Pediatric Acne and Rosacea at the 2026 AAD Annual Meeting: Scarring and Pigment-Change Questions for Pediatric Dermatology Patients