Grades you can audit.
Every claim on this site carries its receipt: the studies behind it, graded by how much they should be trusted. This page is the whole method — short enough to actually read.
The four tiers
Adapted from how evidence is graded in medicine (GRADE and the evidence pyramid), worn lightly. Human trials outrank lab results; independent replication outranks a single striking study.
Strong evidence
Multiple independent randomized controlled trials, or a meta-analysis, agree the ingredient helps this concern. Example: azelaic acid for breakouts.
Good evidence
One or two solid randomized trials, or consistent controlled studies. Real evidence — a thinner file. Example: niacinamide for dark spots at 4–5%.
Early evidence
A single small trial, or mostly lab and animal data. Promising, honest about being unproven. Example: bakuchiol for wrinkles — one 44-person trial.
Not enough evidence
Tradition, marketing, or no meaningful human data. Shown in neutral gray, never a scary red — absence of evidence is not evidence of danger.
What we refuse
No hazard scores
Apps that flash red warnings score ingredients by hazard lists — ignoring dose, concentration, and formulation, which is precisely what toxicologists criticize them for. Hazard is not risk. We refuse the mechanic entirely.
No blanket scores
One number per product hides more than it shows. An ingredient can be excellent for breakouts and irrelevant for wrinkles — so every grade we publish is for one ingredient, for one concern, at studied concentrations.
No pay-to-play
No brand money, no paid placements, no affiliate links in grades. The demo catalog is a hand-curated sample; ingredient lists are factual label data.
Where the data comes from
Ingredients: the European Commission's CosIng database — 24,000+ INCI names, functions, and regulatory annex status (CC BY 4.0).
Studies: Europe PMC, queried per ingredient and concern, filtered to randomized trials, meta-analyses, and systematic reviews, ranked by citations. Every receipt links to the original paper.
Safety context: the EU's scientific safety committee (SCCS), the Cosmetic Ingredient Review (CIR), and the FDA — regulatory conclusions at real-world use, not hazard lists.
Active selection: seeded from a 2025 consensus of 62 cosmetic dermatologists (JAAD Delphi) mapping ingredients to concerns with evidence levels.
Where this demo is honest about itself
Evidence summaries are drafted with AI assistance and were fact-checked claim-by-claim against the primary literature in August 2026 — several grades were corrected in that review, including downgrades where the science was thinner than it first appeared. Independent clinical sign-off is still pending. The product catalog is a labeled sample. Nothing here is medical advice; a dermatologist beats any website.
See it in practice: niacinamide, with receipts →