The Standard
Methodology
Every status on this site traces to a primary regulatory document. This page explains which documents, how we read them, and what we refuse to do with them.
Primary sources
Our statuses are derived from the regulators themselves, not from secondary lists:
- EU cosmetics: Regulation (EC) No 1223/2009 and its Annexes (Annex II, prohibited substances; Annex III, restricted; Annexes IV–VI, colorants, preservatives, and UV filters), checked against the Commission's CosIng database and ECHA listings.
- EU food additives: Regulation (EC) No 1333/2008 and its Union list, plus the amending regulations that change it (for example, Commission Regulation (EU) 2022/63 on titanium dioxide), and the EFSA opinions that prompt them.
- EU chemicals generally: REACH (Regulation (EC) No 1907/2006) and the ECHA candidate and authorization lists.
- US status: the Code of Federal Regulations (Title 21 for food and cosmetics), Federal Register rulemaking, and FDA guidance and constituent updates. Where a state law diverges from federal law (California's AB 418, for instance), we say so explicitly.
How statuses are assigned
Each ingredient carries one of four statuses, and the assignment rule is mechanical, not editorial:
- Banned in EU: the substance appears on a prohibition list (e.g., Annex II of the Cosmetics Regulation) or holds no authorization where one is required. In the EU's positive-list systems, "not authorized" and "banned" have the same practical effect, and we count both.
- Restricted in EU: authorized only under conditions: a concentration cap, a limited product category, or a mandatory warning label. We print the condition, because "restricted" without the detail is just innuendo.
- Legal in both: authorized in both jurisdictions without conditions that meaningfully differ. Most ingredients are in this category, and an honest database says so.
- Banned in US only: the reverse case, where the FDA has acted and the EU has not. Red Dye 3 is the canonical example. We report these with the same prominence; the point is the gap, whichever direction it runs.
What "Restricted" means (and doesn't)
"Restricted in EU" is the most misused phrase in this genre, so we are strict with it. It means the EU permits the substance subject to a written condition, and we cite the condition. It does not mean "almost banned," "being phased out," or "dangerous." Some restrictions are tight (erythrosine: two kinds of cherries). Some are broad (most preservatives carry concentration caps as a matter of course). The citation tells you which kind you're looking at, which is why every entry has one.
Attribution rule
We attribute regulatory conclusions to the regulator that made them: "EFSA concluded," "the SCCS classified," "the FDA determined." We do not convert a regulator's risk language into our own safety language. A flag on this site is a statement about the law, never a statement that a product will harm you.
Update cadence
EU and US regulations are amended continuously. Our working cadence: ingredient statuses are re-checked against ECHA, the EU Official Journal, and the Federal Register on a monthly review cycle, and any page touched by a new amendment is updated and re-dated. Every ingredient page shows its "last reviewed" date. If you find a citation that has drifted out of date, tell us (the contact is on the About page) and we will correct it and note the correction.
Why EWG is not a backbone source
The Environmental Working Group maintains the most widely cited consumer ingredient database in the US, and we read it the way we read everything: as one input. It is not a backbone source for this site, for a specific reason: EWG assigns hazard scores that blend regulatory status, hazard studies of varying quality, and data-gap penalties into a single number. That scoring approach is precisely what this site exists not to do. When we say "restricted," we cite the regulation that restricts; when a study matters, we cite the study. A score with mixed provenance can't be checked. A citation can. Where EWG points us to a primary document we hadn't indexed, we verify the document and cite the document, not the pointer.
Corrections
When we get something wrong, we correct the page, update its review date, and describe the correction. Credibility in this genre is cumulative and fragile; we'd rather print a correction than defend an error.