How we assess

• Clear

  No flagged concerns for your selected context in the sources we reviewed. Not a certification — it means we looked and found no flags.

• Caution

  The evidence is context-dependent or contested: fine for some people or uses, flagged for others. We always tell you which context triggers the caution.

• Flagged

  A major regulator restricts or prohibits this ingredient for this kind of use. We name the authority, quote the finding, and link the source.

• Not yet assessed

  Not in our database yet, or not yet assessed. We never guess — you can request an assessment with one tap.

The product’s verdict is its worst ingredient verdict for your context. A product with unassessed ingredients can’t get a clean Clear — you’ll see Clear plus an “N not yet assessed” badge. Verdicts are computed by fixed rules — no verdict is ever generated or reworded by an AI model.

• EU CosIng. The European Commission’s cosmetic ingredient database and the EU Cosmetics Regulation annexes.
• EFSA. European Food Safety Authority scientific opinions on food additives, including population-specific caveats.
• FDA. US Food and Drug Administration GRAS and SCOGS evaluations.
• PubChem. The US National Library of Medicine’s chemistry database, used to identify substances across their many names.
• Peer-reviewed studies. Published research cited by DOI where a regulator’s position rests on it.

Product lookups by barcode use Open Food Facts, the open product database. The sources don’t endorse BriskGuide — we report what they published, and you can check us.

Whether an ingredient matters depends on who is exposed and how. Phenoxyethanol, a common preservative:

Same ingredient, different question, different answer — and both answers show their source. That’s why the app asks who you’re checking for.

We’ll be straightforward: no external expert has signed this methodology yet. Instead of asking you to trust a name, we built the app so you don’t have to take anyone’s word — ours included.

• We restate; we don’t judge. Every flag restates a regulator’s published position — a regulation annex, an EFSA or SCCS opinion, an FDA evaluation, a national-agency advisory. We don’t flag on our own reading of individual studies.
• Every flag carries the receipt. The source’s finding is quoted in its own words, linked, and archived — one tap from every verdict.
• We check ourselves and publish the result. Each week we re-verify a random sample of assessments against their primary sources and publish the pass rate here.

There’s a “report a problem” button on every verdict. We triage every report within 72 hours. If a verdict is wrong, we correct it and publish the correction in our public changelog — what changed, why, and which source drove it. We don’t quietly edit.

• No black-box scores — the rules are on this page, and every flag shows its source.
• No AI-written safety claims — a model never decides a verdict or writes the explanations you read.
• No advertising. No data sales. Ever.
• Brands cannot pay to change a verdict.
• No medical advice — we summarize published assessments; we don’t diagnose or treat.