AI Photo-Based Calorie Apps Underestimate Meal Energy by About One‑Third

A study conducted by the NIH Clinical Center evaluated four popular photo‑based calorie‑tracking apps—MyFitnessPal, LoseIt!, CalAI and Appediet—using 102 meals prepared in a tightly controlled metabolic kitchen. The meals were weighed precisely, and standardized photos were run through each app. All four apps underestimated the calorie content by roughly 250–345 calories per meal and the fat content by about 30 grams, representing about a one‑third shortfall from the actual values.

The research was part of a larger diet study examining nutrient processing on low‑carbohydrate and standard diets. Researchers noted that while the apps performed better with higher‑calorie meals, they consistently missed lower‑calorie and higher‑fat dishes. The findings suggest users should treat photo‑based estimates as guidance rather than precise measurements, and future tools may need hybrid approaches combining image analysis with manual inputs to improve accuracy.