Statistics Reference
Skin Health Statistics 2026: AI Screening, Usage and UV Data
Every figure below comes from a ScanSkinAI report we have published in full, with the source named and a copy-ready citation attached. Journalists, researchers and writers are welcome to quote these numbers with attribution — no permission needed.
- 95.3%
- Clinically acceptable
- 2%
- Under-triage rate
- 50,265
- Scan records analysed
- 48+
- Countries with UV data
Quick answer
This page collects the statistics ScanSkinAI has published in its own reports, each with the source it came from and a copy-ready citation. Headline figures: 95.3% clinically acceptable accuracy across 80 skin conditions in an independent 100-case dermatologist audit, a 2% under-triage rate, and a retrospective analysis of 50,265 scan-level records. Every figure is free to quote with attribution.
- Last updated
- 2026-08-08
- Version
- 1.0
- Licence
- CC BY-NC 4.0 — free to quote with attribution
AI skin-screening performance
Figures from an independent dermatologist audit of 100 anonymised real-world cases, reviewed case by case in March 2026 across 80 skin conditions.
of AI outputs were clinically acceptable
"Clinically acceptable" means the output would not lead to patient harm or a materially incorrect clinical pathway. It captures both exact agreement and clinically safe partial agreement, and describes screening performance rather than diagnostic performance.
Source: ScanSkinAI Accuracy Report 2026 · As of March 2026
Copy-ready citation
In an independent 100-case dermatologist audit, 95.3% of ScanSkinAI outputs were rated clinically acceptable across 80 skin conditions (ScanSkinAI Accuracy Report 2026, https://www.scanskinai.com/accuracy-report).
under-triage rate
Under-triage means the tool recommended a lower urgency than appropriate — the only triage error that can delay needed care. It occurred in 2 of 100 audited cases.
Source: ScanSkinAI Accuracy Report 2026 · As of March 2026
Copy-ready citation
ScanSkinAI recorded a 2% under-triage rate (2 of 100 cases) in independent dermatologist audit (ScanSkinAI Accuracy Report 2026, https://www.scanskinai.com/accuracy-report).
over-triage rate
Over-triage means the tool recommended a higher urgency than necessary — inconvenient but safe. The audit found the model errs in this conservative direction four times more often than the unsafe one.
Source: ScanSkinAI Accuracy Report 2026 · As of March 2026
Copy-ready citation
ScanSkinAI produced 8 over-triage and 2 under-triage cases out of 100 in independent audit, indicating conservative triage behaviour (ScanSkinAI Accuracy Report 2026, https://www.scanskinai.com/accuracy-report).
skin conditions covered by the audit
The accuracy figure is measured across 80 different skin conditions, not a single-disease benchmark such as melanoma versus benign naevus.
Source: ScanSkinAI Accuracy Report 2026 · As of March 2026
Copy-ready citation
ScanSkinAI's audited accuracy spans 80 distinct skin conditions rather than a single-disease benchmark (ScanSkinAI Accuracy Report 2026, https://www.scanskinai.com/accuracy-report).
Fitzpatrick skin types validated
The audit was stratified across all six Fitzpatrick types — Type I: 7 cases, II: 14, III: 21, IV: 20, V: 13, VI: 5 — deliberately weighted toward Types III–IV rather than validating on lighter skin only.
Source: ScanSkinAI Accuracy Report 2026 · As of March 2026
Copy-ready citation
ScanSkinAI's validation audit was stratified across all six Fitzpatrick skin types, weighted toward Types III–IV (ScanSkinAI Accuracy Report 2026, https://www.scanskinai.com/accuracy-report).
Embed this on your site
Free to reuse in articles and reports. The snippet includes a source link back to this page — please keep it.
<iframe src="https://www.scanskinai.com/embed/ai-accuracy" title="ScanSkinAI audited accuracy figures" width="100%" height="420" style="border:1px solid #e5e7eb;border-radius:16px" loading="lazy"></iframe>
<p style="font-size:12px">Source: <a href="https://www.scanskinai.com/accuracy-report">ScanSkinAI Accuracy Report 2026 — ScanSkinAI</a></p>How people actually use AI skin checks
A retrospective descriptive analysis of historical scan records on the ScanSkinAI platform. These describe platform usage, not population prevalence — the people who scan are self-selecting.
scan-level risk-band records analysed
The report is a retrospective descriptive aggregate analysis with scan records as the unit of analysis, so one person may contribute several records.
Source: Digital Skin Health Insights Report 2026 · As of 2026
Copy-ready citation
ScanSkinAI's Digital Skin Health Insights Report 2026 analysed 50,265 scan-level risk-band records (https://www.scanskinai.com/research/digital-skin-health-report-2026).
medical skin-module AI-output records analysed
These records capture what the medical skin module returned, as distinct from the overall risk band assigned to the scan.
Source: Digital Skin Health Insights Report 2026 · As of 2026
Copy-ready citation
ScanSkinAI's Digital Skin Health Insights Report 2026 analysed 45,270 medical skin-module AI-output records (https://www.scanskinai.com/research/digital-skin-health-report-2026).
distinct AI output labels observed
Roughly 430 distinct labels appeared across the dataset, with the long tail beyond the ten most common accounting for about 26% of scans (≈ 11,900). Skin presentations are far more varied than the handful of conditions that dominate public awareness.
Source: Digital Skin Health Insights Report 2026 · As of 2026
Copy-ready citation
Around 430 distinct AI output labels appeared across ScanSkinAI scans, with the long tail representing roughly 26% of all scans (Digital Skin Health Insights Report 2026, https://www.scanskinai.com/research/digital-skin-health-report-2026).
UV exposure by country
Typical clear-sky peak UV index values compiled from WHO Global Solar UV Index guidance and national meteorological agency published maxima. These are typical seasonal peaks, not live measurements.
countries with published peak UV ranges
The dataset gives a representative city, summer and winter clear-sky peak UV, peak months, and an approximate time-to-redness for unprotected Fitzpatrick I–II skin.
Source: ScanSkinAI UV Index by Country dataset · As of 2026
Copy-ready citation
ScanSkinAI publishes typical clear-sky peak UV index ranges for more than 48 countries (UV Index by Country, https://www.scanskinai.com/blog/uv-index-by-country).
burn-time reference skin types
Time-to-redness estimates in the dataset are given for unprotected Fitzpatrick I–II skin at the local summer peak, the fastest-burning group, so they represent a conservative floor rather than an average.
Source: ScanSkinAI UV Index by Country dataset · As of 2026
Copy-ready citation
ScanSkinAI's UV dataset reports time-to-redness for unprotected Fitzpatrick I–II skin at local summer peak UV (UV Index by Country, https://www.scanskinai.com/blog/uv-index-by-country).
Embed this on your site
Free to reuse in articles and reports. The snippet includes a source link back to this page — please keep it.
<iframe src="https://www.scanskinai.com/embed/uv-index-by-country" title="Peak UV index by country — ScanSkinAI" width="100%" height="620" style="border:1px solid #e5e7eb;border-radius:16px" loading="lazy"></iframe>
<p style="font-size:12px">Source: <a href="https://www.scanskinai.com/blog/uv-index-by-country">UV Index by Country — ScanSkinAI</a></p>How to cite these figures correctly
These rules exist because the most common misquotes are small: dropping the denominator, or turning "clinically acceptable" into "correct". Following them keeps a claim defensible if a reader checks it.
- Name the report, not just the site — for example "ScanSkinAI Accuracy Report 2026" rather than "a study".
- Link to the source report so readers can check the method and limitations for themselves.
- Say "clinically acceptable" rather than "correct" when quoting the 95.3% figure — the two are not the same claim.
- Describe usage figures as ScanSkinAI platform data, never as national or global prevalence.
- Keep the sample size next to the percentage: 2% under-triage means 2 of 100 audited cases.
- Give the as-of date. UV values are typical seasonal peaks, not live measurements.
Questions about this data
Can I quote these statistics in an article or paper?
Yes. The figures are published under CC BY-NC 4.0 and are free to quote in editorial, academic and non-commercial work with attribution. Each entry carries a copy-ready citation, and the page as a whole can be exported as BibTeX or RIS.
Do these numbers describe the general population?
No. The usage figures describe scans submitted to ScanSkinAI by people who chose to check a skin concern, so the group is self-selecting. They should be described as platform data, not as population prevalence or incidence estimates.
What is the difference between under-triage and over-triage?
Under-triage means the tool recommended a lower urgency than appropriate — the only triage error that can delay needed care, recorded in 2 of 100 audited cases. Over-triage means it recommended a higher urgency than necessary, recorded in 8 of 100 cases: inconvenient, but the safe direction of error.
Is 95.3% the same as diagnostic accuracy?
No. It is the proportion of outputs rated clinically acceptable by reviewing dermatologists across 80 conditions, capturing exact agreement plus clinically safe partial agreement. It measures screening performance and is not equivalent to sensitivity or specificity against histopathology.
How often are these figures updated?
Whenever the underlying report is revised. Each entry shows the date the source report was published or last reviewed, and the page header carries a version number and last-updated date so a citation can be pinned.
Cite this page
Ivy AI Solutions Limited. (2026). Skin Health Statistics 2026: AI Screening Performance, Usage and UV Exposure Data. ScanSkinAI. https://www.scanskinai.com/research/skin-health-statistics
These figures describe ScanSkinAI's own reports and platform data. They are not population prevalence estimates, and ScanSkinAI provides preliminary screening information rather than a medical diagnosis.
Full reports behind these numbers
ScanSkinAI Accuracy Report 2026
Independent 100-case dermatologist audit, method and limitations.
Digital Skin Health Insights Report 2026
Retrospective analysis of 50,265 scan-level records.
AI Skin Cancer Screening Methodology
Model design, training, evaluation and explainability.
Research & Evidence Hub
All ScanSkinAI publications, with Zenodo DOIs.
Figures are updated when the underlying report is revised. Last updated 2026-08-08.