Why ingredient checker apps disagree about the same product

Different tools, different questions

Check one label with two tools and one may show a warning where the other shows nothing. Usually neither is broken: they answer different questions, from different sources, in different ways.

They start from different sources

EU cosmetics law works by lists. Annex II of the cosmetics regulation names the substances cosmetics must not contain, with entries numbered up to 1,767 in the consolidated text of 18 May 2026. Annexes III to VI set the conditions for restricted substances, colourants, preservatives and UV filters.¹

US rules work differently. FDA regulations specifically prohibit or restrict a short list, eleven entries on the FDA's own summary. Beyond that list, with an exception for coal-tar hair dyes, any ingredient that makes a cosmetic harmful when used as directed or in the customary way is against the law, whether or not a regulation names it. The FDA's explanation: different countries and regions regulate cosmetics under different legal frameworks.²

Chemical hazard classifications come from a different law again. The EU's CLP regulation classifies substances by the nature of their hazard, and it does not apply to cosmetic products in their finished state.³ Curated lists, drawn up by a retailer, a campaign group or an app, reflect the criteria of whoever compiled them, which is also why clean beauty means different things in different places.

Methylisothiazolinone shows how this plays out. As a pure substance it has an EU harmonised classification as a skin sensitiser, category 1A (CLP Annex VI, index number 613-326-00-9).³ As a cosmetic ingredient, EU law permits it as a preservative in rinse-off products only, up to 0.0015% (Annex V, entry 57).¹ A tool reading the first source flags Methylisothiazolinone as a sensitiser; one reading only the second lists it as permitted within limits. Both are reporting real rules.

Hazard, risk and scores

The EU's Scientific Committee on Consumer Safety keeps two questions apart. Hazard identification asks about a substance's intrinsic properties: whether it has the potential to damage health. Exposure assessment asks how much people meet, from the amount in each type of product and how often it is used.⁴ A label lets a checker report the first but not work out the second, because it carries no amounts. So a tool that flags hazards and one that flags only legal limits can disagree about everyday ingredients. What an ingredient scanner can and cannot tell you works through an example.

Scores widen the gap. Turning a list into one number means deciding how much each flag weighs, whether a name near the end counts for less, and what to do with names nobody matched. Two systems that decide differently will grade one product differently from identical data.

Position is a weak guide to amount. Ingredients run from most to least by weight, but below 1% they may be listed in any order⁵, so a score built on position is estimating quantities the label never states. Tools that decline to score for that reason show a list of facts where others show a grade.

Matching names, and misreading them

Before a checker can say anything, it has to recognise each name, and one ingredient can appear under several, as What is INCI? explains. Names also change: a retired INCI name stays in publication for a transition period, and labels change when existing stock runs out.⁶ The EU replaced its own glossary of label names in 2025, adding new names and correcting ones that had been reported wrongly; the new glossary applies from 30 July 2026.⁷

A tool that misses an alias may call the ingredient unknown or match the wrong one. A photographed label adds misreadings: a zero read as the letter O, two names run together, letters lost on a curved bottle. Whether a tool then guesses, asks or gives up changes what it reports. See how to photograph an ingredient list.

Rules and data change

The consolidated text of 18 May 2026 lists 54 amending regulations since 2013, three of them adopted between May 2025 and April 2026.¹ Change also reaches shelves slowly: when the EU lengthened its list of fragrance allergens that labels must name, products that did not follow the new entries could still be made available until 31 July 2028.⁸

So two checkers with data of different ages can disagree about the same name, each right on the day its data was taken. A tool that dates its sources lets you tell which.

How this site does it

The methodology page sets out the choices behind the ingredient checker and the app, which share one dataset:

• Source. Ingredient names from CosIng, the European Commission's database of cosmetic ingredients, with the names labels commonly use, each ingredient's functions and the EU rules that apply to it. It is selected and normalised, never rewritten, and each source is listed with the date it was retrieved.
• Matching. Each name is recognised, matched as a known alias, corrected, offered as a suggested match or marked unknown. A spelling is corrected only when every difference is a look-alike character, exactly one ingredient fits and no number changes, and it can be undone. Suggestions wait for you to choose, and an unknown name is never treated as harmless.
• No score. No product gets a safety score or a grade, and no ingredient is called toxic, clean, safe or unsafe on its own. A group such as Fragrance is a fact about an ingredient, not a judgement.
• Limits stated. CosIng has no legal value, nothing in the dataset is a safety assessment, and text recognition can misread a label.

What to ask of any checker

• Where does each statement come from? A flag that names its source can be checked.
• Is it about the pure substance or about cosmetics? The two answer different questions.
• Is there a score, and what goes into it? A label gives no amounts, so a score adds someone's judgement.
• What happens to a name it cannot match? Unknown should stay unknown, not become fine.
• How old is the data? Rules and label names change.

If two tools disagree about an ingredient you care about, read the rule behind each flag. If you react to a product, speak to a doctor or pharmacist: no checker can make that call.