How MeSienta analyses a label
MeSienta reads a food label (by barcode, by photo, or by pasting in the ingredients) and hands you back an analysis with plain traffic lights. This page explains where each verdict comes from, with its sources, and also what the app cannot know. If something doesn't match what you see in the app, write to us.
The principle: the AI reads, the rules judge
No MeSienta verdict is decided by artificial intelligence. The AI does two reading jobs and nothing else: pulling text out of photos of a label, and writing explanations. The judgement (high in sugars? an additive with an official warning? high in FODMAPs?) always comes from a rules engine running on your phone, with fixed criteria that can be audited and traced to a source. Two people who scan the same product see the same result.
The nutrition traffic lights: the FSA's public thresholds
The levels shown for sugars, fat, saturates and salt use the thresholds of the UK front-of-pack labelling scheme (Food Standards Agency / Department of Health, the UK government's food standards body), the longest-running traffic-light system in Europe. They are applied per 100 g for solid foods and per 100 ml for drinks. Drinks have their own thresholds too, exactly as the guidance defines them. Thresholds are for salt, not sodium. Where a label declares sodium instead (often in milligrams), the app converts it to salt itself and says so. We don't make up the cut-offs: if a product comes out "high in salt", it is because it goes above the level the FSA calls high.
Additives: official positions only, no scaremongering
The catalogue of flagged additives is built to a strict rule: something gets in only if an official body has ruled on it, namely EFSA (the European Food Safety Authority), IARC (the World Health Organization's cancer research agency) or European law (Regulation 1333/2008 and its amendments). An additive re-evaluated with no safety concerns is not flagged: the absence of a warning is not itself worth flagging. No alarmist lists inherited from blogs.
- High level: solid evidence or an outright ban (the nitrites used in curing, for instance, or titanium dioxide E171, banned in the EU since 2022).
- Medium level: an official warning, an IARC Group 2B classification, or a re-evaluation that left doubts.
MeSienta also runs a check that almost no other app bothers with: it compares the additives the database attributes to a product against the actual text of the ingredients list. If an E number appears in the database record but not on the label, it is marked unverified rather than taken as read. Wrong additive attributions are one of the most criticised failures of nutrition apps, and we would rather admit the doubt than inherit the error.
The catalogue is reviewed monthly against EFSA, IARC and EU regulations. The date of the last review ships in the app's own code.
FODMAPs: the Monash framework, applied to the label
The clinical reference for the low-FODMAP diet is Monash University, which measures generic foods (an apple, a yoghurt). MeSienta does the complementary work: estimating the FODMAP content of a specific supermarket product from its ingredients list, which is what you actually have in front of you on the shelf.
The engine doesn't stop at "it contains onion"; it models what the label declares:
- A "lactose free" claim cancels the lactose from the dairy in that list.
- Garlic-infused oil doesn't carry fructans over: they are not fat-soluble.
- Sourdough fermentation breaks down some of the fructans.
- Order matters: in the EU ingredients are listed in descending order of weight, so a trigger at the end of a long list, or with a tiny percentage, weighs less. Potent triggers at small doses (garlic, onion, inulin, polyols) always count in full.
The dictionaries cover the languages you normally find on European labels (Spanish, English, Portuguese, French, Italian, German and Swedish). And when the text is in a language the engine cannot read, the verdict is an honest "I don't know" — never a false "low in FODMAPs" just because nothing was found.
Where the data comes from
- Open Food Facts, the open, collaborative food database, for product records by barcode.
- USDA FoodData Central as a fallback when a product isn't in Open Food Facts.
- Your own photos of the label, read by AI, when the product is in no database at all or its record is missing data.
- Nutri-Score and NOVA group are shown exactly as Open Food Facts publishes them, as extra context.
What MeSienta cannot know
We would rather tell you ourselves than have you find out:
- Exact amounts. FODMAP content depends on the dose, and an ingredients list doesn't always declare percentages. That is why we talk about an estimate, not a measurement.
- Recent reformulations. If a manufacturer changed the recipe yesterday, the database can take a while to catch up. When in doubt, give it the real label: photos or text.
- Database errors. Open Food Facts is collaborative, and sometimes a record arrives incomplete or out of date. The app tells you when the data it can see isn't enough for a verdict.
- Your own clinical situation. The app does not diagnose and does not treat. The sensitivities you set in your profile tailor how the label is read, nothing more.
Last reviewed: 11 August 2026 · Written by Javier Pino