Methodology

Data sources

SkinCareRadar primarily monitors PubMed (via the NCBI E-utilities API) for newly published research on skincare ingredients, skin concerns, and treatments. Papers are retrieved using targeted search queries for tracked ingredients and concerns, filtered by recency and relevance.

Summarization

Each paper's abstract is processed by a large language model to produce a plain-language summary consisting of: a headline, a short explanation of the findings, and a “why it matters” statement. The model is instructed to maintain scientific accuracy, avoid overstating cosmetic vs. medical outcomes, and flag relevant caveats.

Summaries are generated from the abstract only — we do not access full-text papers. This means complex nuances in the full paper body may not be reflected.

Evidence stage labeling

Each finding is classified into one of eight evidence stages based on the study type described in the abstract:

• in-vitroCell or tissue culture experiments.
• ex-vivoSkin model or tissue experiments (e.g. reconstructed skin).
• animal-studyStudies conducted in animals.
• small-human-studySmall, uncontrolled, or pilot human studies.
• controlled-trialRandomized or controlled human trials.
• systematic-reviewSystematic reviews of published literature.
• meta-analysisStatistical pooling of results across multiple studies.
• guidelineGuidelines or consensus statements from medical societies.

Stage classification is performed automatically and may occasionally be incorrect. If you spot an error, use the feedback form to let us know.

Notable findings

Findings from controlled trials, systematic reviews, or meta-analyses with noteworthy results are tagged as Notable. This is not an endorsement — it indicates a higher evidence level relative to lab or animal studies.

Ingredient and concern classification

Each finding is associated with one or more tracked ingredients and skin concerns via keyword matching against the paper title and abstract. Classification is automated and may occasionally produce incorrect associations, particularly for papers covering multiple topics.

Limitations

• Summaries are generated from abstracts only, not full papers.
• Automated classification may produce errors in stage, ingredient, or concern assignment.
• The system covers a subset of published research — not all relevant papers are indexed.
• AI-generated summaries may miss nuance present in the full paper.
• This site does not constitute medical advice of any kind.