snapkin
The camera · Decision 2 of 12
Why we track added sugar, and what we mean by it
Snapkin says “added sugar” everywhere, because that is the phrase people use. The definition underneath is the World Health Organization's free sugars, which covers more.
Written and maintained by Mattias Geniar. Last updated .
- What the app does
- Counts WHO free sugars with a fruit-purée convention, and uses roughly 10% of your calorie target. No streak on it, no reward for staying under.
- How sure we are Thin , One small study, or an engineering choice.
- Less than anything else the app shows. It is a rough budget, not a measurement, which is why it has no confidence badge.
The evidence for it, and there is little of it
This is the weakest evidence on any of these pages, and the page says so throughout. Here is what there is.
What “added sugar” means here
The definition underneath is the World Health Organization's free sugars: sugar added by a manufacturer, a cook or the person eating, plus honey, syrups, and fruit juice and juice concentrates. It does not include the sugar in whole fruit, in vegetables, or in plain milk and yoghurt. The app also counts fruit purées and smoothies, following the broader processing convention in the 2018 UK definition. WHO’s wording does not explicitly enumerate all these processed-fruit cases. Plain stewed or canned fruit with pieces retained is not automatically free sugar; added sweeteners still count.
The “Added Sugars” line on a US nutrition label (21 CFR 101.9) is not the same thing. Total sugars, the figure a European label declares, is a third thing again. The definitions differ for juice and for sugars naturally present in whole foods.
| Counted as sugar? | Snapkin (WHO free sugars) |
US label “Added Sugars” |
EU label “of which sugars” |
|---|---|---|---|
| Sugar stirred in, or added by a manufacturer | Yes | Yes | Yes |
| Honey, syrups, molasses | Yes | Yes | Yes |
| 100% fruit juice, including correctly reconstituted juice | Yes | No | Yes |
| Concentrated juice used as a sweetener | Yes | Yes* | Yes |
| Sugar in whole fruit and vegetables | No | No | Yes |
| Lactose in plain milk and yoghurt | No | No | Yes |
*US rules contain specific concentrate exceptions. See FDA guidance and the regulation linked above.
The WHO guideline recommends keeping free sugars below 10% of energy, and suggests going below 5% if possible. Snapkin converts 10% of target calories to grams at 4 kcal/g and rounds to the nearest 5 g, so the displayed budget can be slightly above or below 10%. This is a rough tracking budget, not a precisely enforced WHO threshold.
How the number is worked out
Free sugars are not printed on any European label, so they have to be estimated. There is published specialist methods for estimating added and free sugars from food-composition and ingredient information. Our analyser prompt borrows their hierarchy of evidence; an AI following instructions is not the validated specialist procedure. Louie et al. (2015) set out ten steps, starting from hard evidence and ending with guesses by food category. Kibblewhite et al. (2017) restated them for WHO free sugars. Scapin et al. (2021) adapted them for a country whose labels declare neither total nor added sugars.
Scapin checked the method against US label-declared values on 930 products and reached an agreement score (ICC) of 0.98. An independent test of Louie's method (Davies et al., 2022) put it at R² 0.97 with a mean absolute error of 1.26 g per 100 g. Davies used a synthetic test set of 500 Australian products to compare the manual method with a machine-learning method. Both evaluations concern packaged-food information, not photographs.
The reproducibility study asks a different question. Two people applying the same written steps to the same 5,740 foods differed with a standard deviation of 4.6 g per 100 g. That standard deviation cannot be divided by another study’s mean absolute error to claim a multiple of inaccuracy: the metrics, samples and comparisons differ.
The evidence against, and there is a lot of it
The weight effect is largely explained by energy intake
Te Morenga et al. (BMJ 2013): eating less sugar changes weight by −0.80 kg (95% CI −1.21 to −0.39) and eating more by +0.75 kg (0.30 to 1.19). But there was no significant weight difference when sugars replaced other carbohydrates at matched calories. This is a finding about body weight, not proof of no other health effects. WHO’s guideline also considers dental caries.
Estimating sugar from a photograph is the weakest thing this app does
One peer-reviewed photo study reporting total sugar is O'Hara et al. (2025), 114 photographs of weighed meals, assessed with ChatGPT-4. It measured total sugar, not free sugar. We have no validation showing that Snapkin can accurately separate the two from photographs. Hidden ingredients make that a particularly uncertain estimate.
“Carbs that turn into sugar” is a different thing, and we do not track it
Digestible carbohydrate, including starch, can affect blood glucose. Glycaemic load combines a food’s glycaemic index with the amount of available carbohydrate; it is distinct from free-sugar content. Zeevi et al. (Cell 2015) put 800 people on continuous glucose monitors across 46,898 meals and found that different people respond very differently to the same meal. Published glycaemic-index and load tables exist, but do not predict every individual’s response. Snapkin does not estimate your personal blood-glucose response.
So the sugar limit is a rough budget rather than a measurement, and it carries no confidence badge. The eating disorders page has the other half of this decision: there is no streak on the sugar limit, and no celebration for staying under it.