How GLYA estimates calories
Where the figures come from, how the likely range is made, and its limits
Updated October 2026
GLYA gives every meal an estimate. Your day shows it as “about” a number of calories, and each meal’s page shows its likely range, such as 430–610 kcal, with what it was worked out from. This page explains where those figures come from, how the range is made, and what an estimate can’t tell you.
In short
- GLYA works out what you ate (from your words, a photo or a pack’s label) and how much (from what you say, a weight, or the camera).
- The calories come from national food tables, our own checked figures for common foods, the pack’s label or Open Food Facts, never from the recognizer.
- Every meal gets a likely range. The less GLYA knows, the wider it is.
- It’s an estimate: a photo or a sentence can’t weigh food. You can correct any food, and a weight you enter replaces GLYA’s estimate of the amount.
Where the figures come from
- National food tables. More than 23,000 foods from public tables in the US, UK, Canada, France and Japan: USDA FNDDS Survey Foods (2024-10-31), USDA FoodData Central SR Legacy (2018-04), McCance and Widdowson's Composition of Foods Integrated Dataset (2021), Canadian Nutrient File (2026), ANSES-CIQUAL French food composition table (2025), and a selection of about 200 Japanese foods from the Standard Tables of Food Composition in Japan (2020 (8th revised edition), 2023 supplement).
- Common foods. More than 200 common foods use figures we chose and checked by hand. Most are USDA’s; some are from the UK table or are our own approximations from published ranges. These come first. For other foods, when two tables match a food equally well, GLYA prefers them in the order above, and some foods use your region’s table. For foods taken straight from a table, the meal’s page names the table.
- Packs. A nutrition label you photograph is read on your iPhone, and its figures are used for that food. A barcode GLYA doesn’t already know is looked up in Open Food Facts (Open Database License), whose figures are contributed by its community and can be wrong. Only the barcode is sent; see the privacy policy.
- Restaurants. A restaurant’s menu, from its own website once you’ve confirmed the place, or one you scan, only helps GLYA name your dish. The calories still come from the same sources as above. See the privacy policy for what’s sent.
On a meal, each food shows where its figures came from: From the label, Food database or Estimated. In the app, You › About › Food data sources lists each table with its edition and license; the credits are also below.
How an estimate is made
- What the food is. From your words, from a photo (recognized on your iPhone, with Apple Intelligence where it’s available), or from a pack’s label. If your words and the photo disagree, your words win. The recognizer only names foods: it never supplies a calorie or nutrient figure. Calories per gram come from the tables, our hand-checked common foods, a label or Open Food Facts. A few dishes the tables don’t list use our own estimate from a typical recipe. In the rare case GLYA can’t find a food at all, it uses a general figure for a mixed dish, marks the food Estimated, and gives it a much wider range. Pick the right food and the figure is replaced.
- How much. A weight you type, say or weigh is used as given. A count or a household amount (“two slices”, “a bowl”) is turned into grams using typical weights. From a photo, GLYA estimates the portion from the camera: with the LiDAR Scanner on iPhone models that have one; on other models it estimates how much of the table the food covers and assumes a typical height, which is rougher. Otherwise it uses a typical serving, or your usual amount (below). If you ate only part, say so and only that part counts.
- Calories and nutrients. The amount you ate, multiplied by the food’s calories and nutrients per gram, from its table or label. For dishes usually cooked with oil or butter, GLYA adds a typical amount you can’t see, even to a weight you type, and shows it on the meal; it may ask how much was used. When you weigh a dish in GLYA, the oil is part of the weight, not added to it. If GLYA can’t read protein or fat from a pack’s label, or an Open Food Facts record leaves it out, GLYA estimates it from similar foods at the pack’s calories. A missing carbohydrate figure is left unknown, and the meal shows carbs as not known: check the pack itself. Any other nutrient a source doesn’t give is left unknown too: none is ever filled in as zero.
- Drinks. Calories include the alcohol, at the strength of the drink’s table row, such as a typical beer or wine. GLYA doesn’t yet use a strength you give (“5%”, “14%”): drinks are counted at a typical strength, so a stronger-than-usual drink will read low. If GLYA can’t tell how strong a drink is, nothing about its alcohol is saved to Apple Health.
- What it learns from you. Foods you log often are favored when a photo or a phrase is unclear, and after a few weighings of the same food GLYA can use about the amount you usually have (below). If you turn on scale checks, amounts from the camera are adjusted by how they’ve compared with your scale.
How the likely range is made
Each step above can be off, so GLYA carries that uncertainty through instead of hiding it:
- Which food. When GLYA isn’t sure what a food is, it works out the figure for the likeliest food and stretches the range to reach what the other foods it thinks reasonably likely would usually come to. If the food is wrong altogether, the figure can be far off: check the food’s name on the meal and correct it if it’s wrong.
- How much. An amount from a photo, and especially a typical serving, can be further from what was on the plate than an amount you say or weigh, so its range is wider.
- What’s in it. Some foods vary more than a single table figure shows: cooking oil you can’t see in a mixed dish, or the fat in yogurt or milk when you don’t say which kind. The range widens to cover them.
These are combined for each food and then added up for the meal. The “about” figure is GLYA’s central estimate. The likely range shows how far GLYA thinks the true figure could reasonably be from it: wide when the food or the amount is in doubt, narrow when you’ve weighed it. It isn’t always centered on the “about” figure. For a cola that might be diet, or a yogurt that might be full-fat, it reaches down or up to the other kind. For the least certain foods it’s capped, so it can be narrower than GLYA’s real doubt.
When one answer would make the most difference, GLYA may ask one quick question, such as which yogurt it was or how much you ate, and updates the figure and the range from your answer. Weighing is optional. If you do weigh a food, the weight replaces GLYA’s estimate of the amount, so the range narrows. It doesn’t disappear, because what’s in the food can still vary. After a few weighings of the same food, GLYA can use about the amount you usually have instead of a typical serving, when you don’t say how much. The meal says so, and you can switch back.
The range reflects GLYA’s assumptions. It isn’t a hit rate. We haven’t yet compared GLYA’s ranges with weighed real meals, and our own tests so far are simulations built from the same food tables GLYA uses. So we make no claim about how often a meal’s true calories fall inside its range.
Limits
- Every figure is an estimate. A photo or a sentence can’t weigh food.
- If GLYA names a food wrongly, both the figure and its range can be far off. Check the food names on each meal and correct any that are wrong.
- In our own simulated tests, GLYA’s figures tended to come out high rather than low, especially for large portions sized from a photo and before any question was answered. Weighing a food, or giving its weight, removes most of this.
- If you weigh rice, pasta or meat before cooking, say “dry” or “raw” (for example “75 g dry rice”). Otherwise GLYA counts the weight as cooked food, which for dry rice or pasta is roughly a third of the true calories. Oats are the other way round: a weight of oats is usually counted as dry oats, so if you weighed cooked porridge, say “porridge”.
- GLYA can’t see what’s hidden: oil, butter or sugar added in cooking, sauces, or food under other food. Recipes, restaurant portions and packs vary, and the tables give typical values, not your food’s.
- Drinks are counted at a typical strength, not the one on your can or bottle.
- The tables don’t all count the same way. US, Canadian and Japanese figures include fiber in carbohydrate, while UK and French figures don’t, and each table works out calories with its own factors. So the same food can differ by a few percent, more for carbohydrate in foods with a lot of fiber, depending on which table it came from. GLYA doesn’t yet prefer your own country’s table for most foods, so a food’s figures can come from another country’s version of it.
- When GLYA can’t match a food, it uses a rough stand-in, marks it on the meal and doesn’t send it to Apple Health until you confirm or correct it.
- Labels and Open Food Facts can be wrong, and recognition works less well in poor light or on foods it hasn’t learned. On iPhone models without Apple Intelligence it recognizes fewer foods.
- Daily targets are estimates too. They are worked out from your age, sex, height, weight and activity (the Mifflin-St Jeor formula), or from the energy figures in Apple Health, and are never set below a minimum. GLYA doesn’t show a calorie target while you’re pregnant or breastfeeding, or offer weight loss if you’re already at a low weight.
- GLYA is a food log for general wellness. It is not medical advice and is not intended to diagnose, cure, mitigate, treat or prevent any disease or condition. Its calorie and carbohydrate figures are estimates: don’t use them to work out a dose of insulin or any other medicine. Check with your doctor or a registered dietitian before changing how you eat, especially if you are pregnant or breastfeeding, take medication, or have diabetes, kidney disease, an eating disorder or another medical condition. If you have or have had an eating disorder, calorie tracking may not be right for you. GLYA is for adults aged 18 and over.
Counting calories isn’t right for everyone. If food or eating worries you, GLYA lists helplines for your region in You › About › Eating disorder support. If you don’t have GLYA, your doctor can point you to support.
Credits and licenses
- USDA FNDDS Survey Foods (2024-10-31). Public domain. Source: U.S. Department of Agriculture, Agricultural Research Service. FoodData Central: Food and Nutrient Database for Dietary Studies (FNDDS) 2021–2023, released 2024-10-31. Foods selected and converted by GLY LLC; not endorsed by USDA.
- USDA FoodData Central SR Legacy (2018-04). Public domain (CC0 1.0). Source: U.S. Department of Agriculture, Agricultural Research Service. FoodData Central: SR Legacy, released 2018-04. Foods selected and converted by GLY LLC; not endorsed by USDA.
- McCance and Widdowson's Composition of Foods Integrated Dataset (2021). Open Government Licence v3.0. Public Health England (2021). McCance and Widdowson's The Composition of Foods Integrated Dataset 2021. Contains public sector information licensed under the Open Government Licence v3.0.
- Canadian Nutrient File (2026). Open Government Licence – Canada. Contains information licensed under the Open Government Licence – Canada. Source: Health Canada, Canadian Nutrient File. Values converted to GLYA's units and foods selected by GLY LLC; not endorsed by Health Canada.
- ANSES-CIQUAL French food composition table (2025). Licence Ouverte / Etalab 2.0. Source: Anses. 2025. Ciqual French food composition table. https://ciqual.anses.fr/ (last updated 2025-11-03).
- Standard Tables of Food Composition in Japan (2020 (8th revised edition), 2023 supplement). Public Data Terms of Use v1.0 (公共データ利用規約, compatible with CC BY 4.0). 出典:日本食品標準成分表(八訂)増補2023年. Source: Ministry of Education, Culture, Sports, Science and Technology (MEXT), Standard Tables of Food Composition in Japan 2020 (Eighth Revised Edition), 2023 supplement, processed by GLY LLC: food names translated into English, foods selected, app categories assigned, and only some nutrients carried. Not a MEXT publication.
- Open Food Facts: product figures are available under the Open Database License (ODbL) 1.0. Figures are contributed by its community and can be wrong.
GLYA and GLY LLC aren’t affiliated with or endorsed by any of these organizations, and nothing here is presented as an official publication of theirs.
Try it
Log up to 3 meals free on your first day, so you can see their ranges for yourself. Then, if you’re new to the yearly plan, try it free for 7 days and log as many meals as you like; we remind you two days before the trial ends, in the app and with a notification if you allow them. After it, the yearly plan is $49.99 a year unless you cancel at least 24 hours before it ends (Settings › your name › Subscriptions). GLYA on the App Store · About GLYA · GLYA support