ScanSkinAI Logo
Scan Skin
by Ivy AI

ScanSkinAI Research Report

Digital Skin Health Insights Report 2026

A descriptive scan-level analysis of ScanSkinAI risk bands and AI-supported medical skin-output patterns.

Report information

Digital Skin Health Insights Report 2026 — A Descriptive Scan-Level Analysis of ScanSkinAI Risk Bands and AI-Supported Medical Skin-Output Patterns

Author:
Dr Lifeng (George) Zhu
Affiliation:
Ivy AI Solutions Limited
Publication date:
29 July 2026

Read the Digital Skin Health Report 2026 methodology for the dataset scope, denominators and limitations behind these figures.

Publication edition
2026
Data coverage
1 January 2026 – 28 July 2026
Data cut-off
28 July 2026
Version
1.0
Report type
Retrospective descriptive aggregate analysis
Unit of analysis
Scan records
Risk-band dataset
50,265 scans
Medical skin-output dataset
45,270 scans

Free to download for research, education and non-commercial reference. Commercial use is not permitted.

Important: This report analyses scan records, not confirmed diagnoses or unique individuals. A person may have submitted more than one scan. AI-supported output labels and risk bands are preliminary screening information and must not be interpreted as clinical diagnoses, disease prevalence or diagnostic-accuracy results.

How to interpret this report

This report is provided for research transparency and general educational purposes. ScanSkinAI outputs are preliminary screening information and are not medical diagnoses. A low-risk or normal-skin output cannot rule out disease. Anyone concerned about a new, changing, painful, itching, bleeding, crusting or non-healing skin mark should seek advice from an appropriately qualified healthcare professional.

Executive summary

This report presents a retrospective descriptive analysis of approved aggregate ScanSkinAI scan-history data. It uses approved aggregate historical scan data collected through routine use of the ScanSkinAI platform. The unit of analysis is a scan record, not a person.

Two separate analysis datasets are described. The risk-band distribution included 50,265 scan records. The AI-output distribution included 45,270 medical skin-module scans. These are different datasets with different denominators and must not be combined or presented as the same population. The AI-output distribution was available for a narrower eligible subset of medical skin-module scans. No further record-level breakdown is published.

Data coverage: 1 January 2026 to 28 July 2026. The analysis covers eligible production outputs included in the approved aggregate extract. Model-version-level results are not reported publicly.

Within the medical skin-module subset, the most frequent individually named AI-supported output category was benign naevus / mole at 13.9% of module scans. “Normal skin” represented 11.8% and contact dermatitis 7.3%. Approximately 26% of medical skin-module scans were distributed across approximately 430 other lower-frequency labels, showing that the output distribution extends well beyond the most frequent named categories.

Within the risk-band dataset, 52.8% of scan records fell in the 1–39 low band, 27.7% in the 40–69 moderate band and 18.6% received a 70–100 high-concern flag. A small proportion recorded a score of zero (0.7%) or had no reportable risk-band value (0.3%).

These figures describe platform activity only. A high-concern flag does not confirm cancer, and a low score does not rule out a medical condition. The data cannot be used to estimate disease prevalence or incidence, to measure diagnostic accuracy, or to describe clinical outcomes. No individual user records, identifiers or images are published.

Full definitions, statistical methods, privacy controls and limitations are set out in the methodology.

Two datasets, two denominators

Risk-band dataset

50,265

scan records

Used for every risk-band percentage on this page.

Medical skin-module dataset

45,270

scans

Used for every AI-output percentage on this page.

The AI-output distribution was available for a narrower eligible subset of medical skin-module scans. No further record-level breakdown is published.

Risk-Band Distribution

Risk bands are platform output categories applied to 50,265 scan records. They are not a clinical staging system, a cancer diagnosis or a validated population-risk category.

  • 70–100 — 18.6%
  • 40–69 — 27.7%
  • 1–39 — 52.8%
  • 0 — 0.7%
  • Missing — 0.3%

Risk-band distribution across 50,265 scan records. Percentages are rounded and may not sum to exactly 100%.

  • 70–100

    Category
    High concern flag
    Scan records
    9,323
    % of 50,265 scans
    18.6%

  • 40–69

    Category
    Moderate
    Scan records
    13,904
    % of 50,265 scans
    27.7%

  • 1–39

    Category
    Low
    Scan records
    26,538
    % of 50,265 scans
    52.8%

  • 0

    Category
    Score zero
    Scan records
    353
    % of 50,265 scans
    0.7%

  • Missing

    Category
    No reportable risk-band value in the approved extract
    Scan records
    147
    % of 50,265 scans
    0.3%
A high-concern flag is a screening output, not a confirmed cancer diagnosis. A low score does not rule out disease. These figures are not cancer rates and are not disease prevalence.

Most Frequent Medical Skin-Module AI Outputs

The AI-output analysis included 45,270 medical skin-module scans. Output labels represent preliminary model-generated categories and not clinician-confirmed diagnoses.

Percentage of 45,270 medical skin-module scans

  • Benign naevus / mole13.9%
  • Normal skin11.8%
  • Contact dermatitis7.3%
  • Basal cell carcinoma indication6.5%
  • Acne vulgaris5.9%
  • Melanoma indication5%
  • Vascular lesion3.7%
  • Folliculitis3.5%
  • Seborrhoeic keratosis3.1%
  • Molluscum contagiosum2.4%

† Indication means a preliminary AI-supported output category requiring appropriate interpretation. It is not a confirmed diagnosis.

Complete AI-output table

Approved AI-supported output groups across 45,270 medical skin-module scans. Approximate values are marked ≈.

  • Benign naevus / mole

    Scan count
    6,270
    % of module scans
    13.9%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Normal skin

    Scan count
    5,336
    % of module scans
    11.8%
    Interpretation note
    Does not guarantee the absence of disease.

  • Contact dermatitis

    Scan count
    3,309
    % of module scans
    7.3%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Basal cell carcinoma indication

    Scan count
    2,932
    % of module scans
    6.5%
    Interpretation note
    Concern indication only; not confirmed basal cell carcinoma.

  • Acne vulgaris

    Scan count
    2,688
    % of module scans
    5.9%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Melanoma indication

    Scan count
    2,273
    % of module scans
    5%
    Interpretation note
    Concern indication only; not confirmed melanoma.

  • Vascular lesion

    Scan count
    1,683
    % of module scans
    3.7%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Folliculitis

    Scan count
    1,593
    % of module scans
    3.5%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Seborrhoeic keratosis

    Scan count
    1,402
    % of module scans
    3.1%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Molluscum contagiosum

    Scan count
    1,081
    % of module scans
    2.4%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Dermatofibroma

    Scan count
    984
    % of module scans
    2.2%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Eczema / atopic dermatitis

    Scan count
    835
    % of module scans
    1.8%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Tinea corporis

    Scan count
    825
    % of module scans
    1.8%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Impetigo

    Scan count
    796
    % of module scans
    1.8%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Squamous cell carcinoma indication

    Scan count
    Not publicly reported
    % of module scans
    1.2%
    Interpretation note
    Concern indication only; not confirmed squamous cell carcinoma.

  • Actinic keratosis

    Scan count
    516
    % of module scans
    1.1%
    Interpretation note
    Preliminary AI-supported output; not a clinical confirmation.

  • Not a skin image / unclassified

    Scan count
    305
    % of module scans
    0.7%
    Interpretation note
    The submitted image could not be assigned to a reportable skin-output category.

  • All other labels

    Scan count
    ≈ 11,900
    % of module scans
    ≈ 26%
    Interpretation note
    Approximately 430 lower-frequency output labels grouped for readable aggregate reporting.

The exact approved count for squamous cell carcinoma indication was not included in the supplied aggregate extract. It is shown as a percentage only; no count has been estimated or reverse-calculated.

A Broad Long Tail of Output Categories

≈ 26% of the 45,270-scan module dataset (≈ 11,900 scans) fell across ≈ 430 other labels.

  • The grouped category shows the output distribution extends beyond the most frequent named labels.
  • It does not establish the prevalence of approximately 430 diseases.
  • Some labels may be rare.
  • Category frequencies may be affected by the platform’s users, use cases, model taxonomy and repeated scans.

What the findings may indicate

  • Benign naevus / mole was the most frequent individually named AI-output category in the approved extract.
  • The output distribution includes both common inflammatory concerns and lesion-related indication categories.
  • A substantial proportion of output records is distributed across a long tail of lower-frequency labels.
  • More than half of risk-band scan records were in the low band.
  • Moderate and high-concern bands remain screening outputs requiring responsible next-step communication.
  • The distribution may help inform product transparency, user education and future research planning.

What This Report Cannot Show

  • The number of unique people represented
  • Disease prevalence
  • Disease incidence
  • Confirmed diagnoses
  • Confirmed cancer cases
  • Diagnostic accuracy
  • Sensitivity or specificity
  • Clinical outcomes
  • Whether users followed medical advice
  • Whether repeated scans came from the same concern
  • Geographic differences
  • Age-related differences
  • Sex or gender differences
  • Skin-tone differences
  • Seasonal patterns
  • Changes over time
  • Causes of the observed distribution
  • General-population risk

Privacy and Responsible Aggregate Reporting

  • Only approved aggregate statistics are published.
  • No names, emails, telephone numbers, IP addresses or direct identifiers are displayed.
  • No raw skin photographs or facial images are published.
  • No row-level records are made publicly available.
  • No pseudonymous user identifiers are published.
  • Small or sensitive individual subgroups are not displayed.
  • The report is not used to make employment, insurance or healthcare decisions about named individuals.
  • Access to underlying production systems remains restricted.
  • The public report does not allow readers to trace a statistic to an identifiable person.

The public report contains aggregate statistics designed not to identify individual users.

Limitations

  1. Limitation 1

    Scan-level denominator

    A person may contribute more than one scan, so counts describe scans and not people.

  2. Limitation 2

    Unknown unique-user count

    The approved aggregate data does not establish how many unique people are represented.

  3. Limitation 3

    Self-selected platform population

    People using ScanSkinAI are not a random sample of the general population.

  4. Limitation 4

    AI-output status

    Output labels are preliminary model-generated categories, not confirmed diagnoses.

  5. Limitation 5

    No clinical ground-truth analysis

    This report does not compare outputs with dermatologist-confirmed diagnoses.

  6. Limitation 6

    No accuracy evaluation

    The report must not be used to infer sensitivity, specificity, predictive value or diagnostic accuracy.

  7. Limitation 7

    Different analysis denominators

    Risk-band and AI-output results were generated from different eligible scan subsets.

  8. Limitation 8

    Repeated use

    Repeat scans may influence category counts.

  9. Limitation 9

    Limited public variables

    Age, geography, skin type, device, sex, symptoms and follow-up outcomes are not included.

  10. Limitation 10

    Historical product changes

    Model, interface, taxonomy or threshold changes may influence output patterns if more than one version is represented.

  11. Limitation 11

    Approximate values

    “All other labels” figures are approximate and percentages are rounded.

  12. Limitation 12

    Long-tail grouping

    Combining approximately 430 labels reduces detail but avoids an unwieldy and potentially misleading table.

  13. Limitation 13

    No population inference

    Results cannot be converted into national or global disease rates.

  14. Limitation 14

    No causal conclusions

    The report describes distributions only.

Related research

Authors, review and disclosure

Data coverage start date
1 January 2026
Data cut-off date
28 July 2026
Data extraction date
29 July 2026
Publication date
29 July 2026
Report author
Dr. Lifeng Zhu, PhD — Founder, ScanSkinAI
Data-analysis owner
Dr. Lifeng Zhu, PhD — Founder, ScanSkinAI
Clinical reviewer
ScanSkinAI Clinical Review Team — reviewed 29 July 2026
Technical / data reviewer
ScanSkinAI Data & Engineering Team — reviewed 29 July 2026
Privacy / governance reviewer
ScanSkinAI Privacy & Governance Function — reviewed 29 July 2026
Version
1.0
Next review date
29 July 2027

Funding

This report was prepared and funded by ScanSkinAI unless otherwise stated.

Conflict of interest

ScanSkinAI developed and operates the technology from which these aggregate statistics were generated. This is a company-authored descriptive report. Commercial involvement and the limitations of company-controlled platform data should be considered when interpreting the findings.

Corrections

To request a correction, email info@scanskinai.com. Substantive corrections are versioned and recorded in the methodology’s version history.

Cite this report

Zhu, L. (2026). Digital Skin Health Insights Report 2026. Ivy AI Solutions Limited. https://doi.org/10.5281/zenodo.21669080

This report is archived on Zenodo under the DOI above. DOIs belonging to other ScanSkinAI publications must not be used to cite it.

Download & permitted use

This report is made available for research transparency, education and non-commercial reference only. It must not be used for any commercial purpose, including advertising, resale, marketing claims, investor materials or product promotion, without prior written permission from ScanSkinAI. Please cite the DOI when referencing the report.

This report is provided for research transparency and general educational purposes. ScanSkinAI outputs are preliminary screening information and are not medical diagnoses. A low-risk or normal-skin output cannot rule out disease. Anyone concerned about a new, changing, painful, itching, bleeding, crusting or non-healing skin mark should seek advice from an appropriately qualified healthcare professional.

This report is provided for research transparency and general educational purposes. ScanSkinAI outputs are preliminary screening information and are not medical diagnoses. A low-risk or normal-skin output cannot rule out disease. Anyone concerned about a new, changing, painful, itching, bleeding, crusting or non-healing skin mark should seek advice from an appropriately qualified healthcare professional.

Read the full methodology →