What the research found

Researchers evaluated four popular photo-based calorie-tracking apps (MyFitnessPal, LoseIt!, CalAI, and Appediet) against meals prepared in a metabolic kitchen with precisely measured ingredients. Across all four apps, calorie estimates fell short by 250–345 calories per meal on average, and fat was underestimated by roughly 30 grams. The apps were somewhat more consistent when assessing carbohydrates than other macronutrients.

When tested on a larger sample of over 200 additional meals, the apps appeared to struggle more with ketogenic-style foods—which are typically higher in fat. This suggests the underestimation of calories may be partly driven by systematic difficulty recognising and quantifying fatty foods. Two of the apps (MyFitnessPal and LoseIt!) showed better accuracy on higher-calorie meals compared to lower-calorie ones, hinting that portion-size estimation may play a role in the error.

Why it matters for you

If you're using a photo-based app to track intake for body composition work, strength training, or any protocol where caloric or macronutrient precision matters, expect your actual consumption to be significantly higher than what the app reports—particularly for fat intake. For someone on a higher-fat protocol (carnivore, keto, or even a moderate low-carb approach), this gap could translate to 250–345 uncounted calories per meal, which compounds quickly.

This is relevant whether you're managing intake for a cut, optimising fuel for training, or tracking macro ratios for hormone support or performance. The discrepancy is large enough to throw off calculations for caloric deficit or protein-to-energy ratios. The most practical fix: occasionally weigh portions or cross-check app estimates against a reference like the USDA database or a scale-based tracker, especially for fat-heavy meals. Hybrid tracking—photo plus manual portion entry—will be more reliable than photos alone.

Caveats

  • Preliminary findings: Results presented at a conference abstract stage; not yet peer-reviewed or published in a journal
  • Limited meal set: Initial validation used only 102 standardized meals from a research kitchen; real-world meals vary far more
  • App-specific variability: Performance differed between apps, but sample size per app not disclosed
  • Controlled environment bias: Metabolic kitchen meals may not reflect typical real-world plating, lighting, or photography angles users encounter
  • Ketogenic meals caveat: The preliminary indication that low-carb meals were harder to assess is based on a subset of the 200+ additional meals; effect size and statistical rigor not detailed