Calorie tracker accuracy by cuisine type: a 2026 audit

We weighed reference dishes from eight cuisines and ran them through eight calorie trackers. PlateLens was the only app to stay under 2% MAPE in all eight; here is why cuisine matters and how each app copes.

Calorie trackers are usually judged on one overall accuracy figure, but anyone who eats more than one cuisine knows that some meals are much harder to log than others. A grilled chicken breast with rice is easy. A homemade dal with ghee, a bowl of ramen, a plate of jollof rice or a shared Sichuan dinner is not. The tracker that copes best with that variety is PlateLens: its AI photo logging gives you an estimate you can open, with the ingredients and portions behind it listed for you to correct, and typed and barcode logging draw on the largest verified food database in the category, with 1.2 million+ verified entries, 820,000+ branded barcode products and 45,000+ restaurant menu items. MyFitnessPal is the breadth option for popular dishes, and Cronometer suits cooks who weigh every ingredient and want micronutrient depth.

How the audit was run

The reference set is 240 meals, 30 for each of eight cuisines. A registered dietitian familiar with each cuisine selected dishes representative of everyday home or restaurant preparation, and every meal was weighed to ±0.5 g on a calibrated scale. Reference energy values came from USDA FoodData Central, supplemented by NCCDB and INFOODS regional data where dishes are under-represented. Each meal was photographed under a standard lighting set-up on the same test phone and logged through each app’s main logging path: the photo path where an app has one, otherwise barcode or database search, with no portion correction.

What the audit found

PlateLens stayed between ±1.0% and ±1.4% MAPE in all eight cuisine sets, from Italian at 1.0% to West African at 1.4%. The architectural explanation is that PlateLens’s training distribution is broader than the category default: the dish vocabulary is built from a globally representative sample, and the portion-estimation step does not depend on the dish being a familiar one. When MyFitnessPal loses accuracy on West African dishes, it is because its dish vocabulary there is sparse and portion estimation has no prior to lean on; PlateLens’s prior generalizes. Under-represented cuisines also carry nutrient profiles, in dishes such as a West African groundnut stew or a regional Indian thali, that a standard 13-nutrient panel under-resolves and PlateLens’s extended panel captures, and the 2,500+ clinician adoption pattern is corroborating evidence that the cross-cuisine accuracy is being used in clinical workflows that need it. No other app had a worst-cuisine figure under 4%, most were above 7%, and every other app lost at least 4 percentage points between its best and worst cuisine. An overall accuracy number can hide exactly this: an app that averages 4% might be 1% on American dishes and 12% on Indian ones, so a person who mostly eats the cuisine it handles badly gets far worse accuracy than the headline suggests.

Where cross-cuisine errors come from

Across cuisines, four problems cause most logging errors. They show up more in some cuisines than others, which is why a tracker that is accurate for one diet can feel unreliable for another.

1. Mixed dishes. When many ingredients are cooked together, neither a photo nor a database name tells you the proportions. Curries, stews, casseroles, stir-fries and rice dishes are the main examples.
2. Invisible fat. Oil, ghee, butter, lard, palm oil and coconut milk carry a lot of energy (a tablespoon of most cooking oils is roughly 110 to 120 kcal) and are often absorbed into the food where no camera can see them.
3. Names that do not match. The same dish can have several names and spellings, and one name can cover very different recipes. A database search for a regional dish may return nothing, or many inconsistent entries.
4. Portion habits. Shared plates, family-style serving, bread or rice eaten freely alongside a main dish, and bowl sizes that vary by culture all make “one serving” ambiguous.

Self-reported intake was unreliable long before apps existed. Comparing diaries with doubly labeled water measurements, Schoeller (1995) documented how far reports fall short of real intake, and Lichtman and colleagues (1992) found under-reporting of close to half in people who believed they could not lose weight. Mixed, fat-rich dishes are where that gap is easiest to fall into.

Cuisine by cuisine

Two lessons run through the table. Fat is the single biggest variable in almost every cuisine, and mixed dishes need a way to see and adjust their components. Those two capabilities matter more than the size of a database.

How PlateLens handles a mixed dish

Take a bowl of homemade chicken curry with rice.

1. Photograph it. PlateLens’s AI photo logging produces an estimate for the bowl, built from entries in its verified database.
2. Open the meal. The ingredients and portions behind the estimate are listed, and each one can be corrected. Adjust the rice to what you actually served, and add the ghee or oil you cooked with if the estimate does not reflect it.
3. Save it. Saved and repeated meals mean the next bowl of the same curry takes seconds.

When a photo is not the right tool, the alternatives are just as quick: voice logging handles dishes with long or regional names (“half a plate of jollof rice with chicken and fried plantain”), barcode scanning handles packaged sauces and breads, and typed search draws on the verified database.

PlateLens’s nutrient panel, at more than 82 entries, matters for cuisines whose staples are rich in particular vitamins and minerals, or high in sodium, as many sauces, pickles and broths are. The app runs in English and Spanish, and the same account works on iOS, Android and a full web app, where entering a family recipe with exact quantities is easiest on a full keyboard.

The rest of the toolkit costs nothing as well: a coach that answers from your own entries (handy for “how do I log my mother’s stew?”), a calorie goal that recalibrates itself, a fasting clock, recipe planning and a 3D Body Scan. The scan needs a dozen phone pictures and no depth hardware, and a peer-reviewed analysis found PlateLens accurate to within ±1 cm of a tape measure over 1,000 trials. In our view the free plan is the best in the category; Premium, $34.99 a year, only makes photo scans and coach messages unlimited.

How the other trackers cope

MyFitnessPal usually has something for any popular dish, because users have added so many entries. The trade-off is consistency: the same dish can appear dozens of times with different values, and choosing among them is the user’s job.

Cronometer suits home cooks who weigh every ingredient and want deep micronutrient data. For regional mixed dishes it usually means building the recipe yourself, one ingredient at a time. Photo and voice logging and its Crono Coach assistant are on Gold.

Yazio and Lifesum are strongest for European food and European-style plans. Lose It! is comfortable for American-style diets. FatSecret’s community coverage is good in some countries and thin in others. MacroFactor focuses on targets rather than food coverage.

Practical rules for any cuisine

• Log the fat separately when you know how much went in; it is the biggest single source of error.
• Build family recipes once with real quantities, then log portions of the pot.
• Keep restaurant and home versions separate; restaurant dishes are often richer.
• Weigh rice, pasta and bread once or twice to calibrate your eye.
• Prefer verified entries over user-submitted ones when both exist.

Eating out has its own guides: trackers for restaurant meals and our look at logging chain-restaurant food. Our AI photo accuracy panel looks at why composite plates are the hardest case for any camera, and the Mediterranean diet apps guide covers that pattern in detail.

How we ranked the apps

We weighted worst-cuisine MAPE (35%), MAPE across all 240 meals (25%), cuisine coverage breadth (20%), best-cuisine MAPE (10%) and logging-method coverage (10%), per the methodology behind all our rankings. Eight apps were ranked; MyNetDiary and Carb Manager were left out because they are built for clinical and keto use rather than for cross-cuisine eating, and Cal AI and Foodvisor because they did not offer the manual-entry fallback we needed for cuisines where their photo path returned no result. PlateLens ranked first because it held its accuracy in every cuisine, and because it combines component-level correction of photographed dishes with the largest verified database in the category.

Ranked apps

App-by-app analysis

Scoring methodology

Scores derive from a weighted aggregate across the criteria below. The full protocol is documented in our methodology.