How Label X-Ray scores products

Last updated: October 2, 2026

What goes in

• The ingredient list. From a barcode or name search it comes from Open Food Facts (food) or Open Beauty Facts (cosmetics), free community-maintained databases. Or you paste it, or photograph the label and the app reads it.
• The product name, for context.
• Your allergies and diets, if you set them, so they can be checked against the same list.

Nothing else. There is no price, no brand reputation, and no sales data in the analysis.

How the verdict and score are made

Label X-Ray doesn't use a fixed point-deduction formula. The AI model reads the ingredients and first picks a verdict:

• Clean — the ingredients look minimally processed, with no flagged concerns. Our own shorthand, not a certification.
• Greenwashed — the label's marketing (like “natural”) doesn't line up well with what the ingredients actually show.
• Processed — the ingredients show meaningful processing, or concerns worth knowing about.

Then it scores within a range for that verdict, based on how many concerns it flags, how serious they are, and any genuine positives it finds. The score is labeled for quick reading:

The result lists the specific ingredients that moved the score, worst first, each with a plain-language reason. There is also a “Good stuff” list for genuinely beneficial ingredients, and a “What to do” suggestion for a better choice.

The kinds of flags

Not every flagged ingredient is a warning. Each flag carries a type, so a harmless thickener isn't presented like a real concern:

• Established concern — a well-evidenced safety or health concern. Used sparingly.
• Consider based on your needs — matters for some people, not everyone (for example high sodium or sugar).
• Common functional ingredient — sounds industrial but has no real evidence of harm at normal use. Most “why is this here?” ingredients land here, on purpose.
• Environmental consideration — about sourcing or sustainability, not personal health.
• Insufficient information — the research is mixed or too limited to say either way.
• Selected allergen — matches something you told the app to watch for.

What is not in the score

• Nutrition facts. The Nutri-Score, NOVA group, and per-100 g numbers come straight from Open Food Facts and are shown separately. They do not change the verdict or score.
• Diet checks (vegan, gluten-free, and the rest) are answered separately from the ingredient list and aren't part of the score either.
• Allergy alerts are shown at the top of the result so they can't be missed. A matching ingredient also appears in the flags list, marked as a selected allergen.
• Better options. The like-for-like suggestions on a food result are ranked with Open Food Facts' own Nutri-Score and NOVA data, not with Label X-Ray's score. They are real listings from that database; a small AI step only removes ones that aren't true substitutes, and can't add products that aren't there. Tapping one gives it a full Label X-Ray result.
• The organic indicator is a reading of the ingredient list, not a legal organic certification.

Where the other data comes from

• Product listings and nutrition: Open Food Facts and Open Beauty Facts, under the Open Database License.
• Recalls: searched live in official government databases (US CPSC and FDA, Health Canada, EU Safety Gate and RASFF), always linking to the official notice.
• The ingredient guide: each page is checked against primary regulator sources (such as the FDA, Health Canada, and the EU), and compares how each region treats the ingredient.

Independence

• No brand can pay for a better result. Nothing in the app lets a brand buy a score, a ranking, or a mention, and the code has no brand-specific handling. Each result is generated from the ingredient list in front of it.
• No ads, and no data sold. The app and this site run no ads or ad trackers, and we don't sell personal data.
• No affiliate links. The “better choice” suggestion links to a plain image search, not a store.
• Not affiliated with any brand. Label X-Ray isn't sponsored or endorsed by any brand or manufacturer it mentions (see our Terms).
• How we're funded: subscriptions and one-time scan packs from people who use the app (see Support for plans). Recall search is always free.

Where it can be wrong

• It's an AI, so it can be wrong, incomplete, or out of date. Formulations change, and research moves.
• Scores can vary a little between runs. The same ingredients can land a few points apart on different days. The verdict and the flagged ingredients are usually stable.
• An empty flags list is not a safety guarantee. It means nothing we checked crossed the bar to flag.
• Community listings have errors. A barcode can show the wrong product, a missing ingredient list, or the version sold in another country. The same product often has different ingredients in different countries.
• Diet checks read the ingredients only. They don't know about certifications or cross-contact in the factory.
• Results are opinions, not facts. They're an AI's assessment of public ingredient information, not a verified claim about any manufacturer.
• Not medical advice. For an allergy or a medical condition, read the actual package and ask a professional.

Questions