How we evaluate
every ingredient.
Ingreview checks each ingredient against five evidence layers — from canonical identifiers all the way to community-reported sensitivities. Here's exactly how that works.
The evidence ladder
Five layers, evaluated in order. Each layer adds signal; none replaces the others, and none of them is a guarantee of safety.
INCI database — canonical ingredient identifiers
Every product label is mapped to standardised INCI (International Nomenclature of Cosmetic Ingredients) names. This normalisation step is the foundation of everything else: it means "Retinol", "Vitamin A", and "all-trans-Retinol" all resolve to the same entry, enabling consistent cross-database lookup regardless of how a brand writes its label.
Cosmetic safety assessments
CosIng (EU) and SCCS opinions form the baseline safety tier for cosmetic ingredients. These assessments aggregate years of regulatory and scientific review into structured records. Ingreview's ingredient dictionary is pre-populated from these sources, with updates applied as sources change — see the update-frequency table below for current cadence. Each ingredient's entry reflects the current consensus, including any disagreements between sources, rather than a single flattened number.
Regulatory lists
EU Annex II–VI restricted and prohibited lists are checked automatically on every scan, against Ingreview's ingredient dictionary — populated from CosIng, with updates applied as sources change rather than queried live at scan time (see the update-frequency table below for current cadence). An ingredient appearing on any of these lists is flagged immediately, with the specific regulation cited. Where a restriction applies only above a certain concentration, that context is shown alongside the flag rather than being hidden inside a rating.
Peer-reviewed studies
PubMed-indexed research is linked where available. Ingreview does not summarise study conclusions without citing the source — if we can't link to a primary paper, we say so.
Community reports
User-flagged sensitivities and reactions are shown separately and never mixed into the risk score. This separation is intentional: an anecdotal reaction, however real, carries different evidentiary weight than a regulatory prohibition. You can see community signals alongside the scientific evidence, but the two are always clearly labelled so you can apply your own judgement.
Alongside the layers above, each ingredient entry reports which reference categories the catalogue actually holds for it — regulatory status, scientific opinion, literature, community reports. That is a statement about our record, not a verdict on the ingredient, and no combination of layers ever means an ingredient is absolutely safe: ingredient safety is context-dependent, dose-dependent, and varies by individual. For Ingreview's full wording policy and how these signals are kept separate, see our Data Sources page.
Data update frequency
How often each layer is refreshed. All figures are targets for the v1.0 launch; current pre-launch cadence may differ.
| Layer | Source | Target refresh | Status |
|---|---|---|---|
| INCI database | CosIng EU | Quarterly | planned |
| Safety assessments | CosIng, SCCS | Monthly | planned |
| Regulatory lists | EU Annex II–VI | On publication | planned |
| Peer-reviewed studies | PubMed | Weekly index pull | planned |
| Community reports | User submissions | Real-time | planned |
Privacy and your data
Your scan history, profile, and personal sensitivity flags are processed on-device and never shared without your consent. For the full picture, see our Privacy & Terms page.