Four of the most popular AI-powered calorie-tracking apps routinely undercount the energy in a single meal by 250 to 345 calories — a gap large enough to cancel out a brisk 45-minute walk. The finding, from researchers at the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), was presented at NUTRITION 2026 on July 25 and carries direct implications for anyone relying on photo-based logging to manage their diet — a topic covered in depth at nutritionist.today.
What the Study Found
Scientists led by Aaron Hengist photographed 102 meals prepared in a tightly controlled metabolic kitchen, where every ingredient was weighed to 0.1-gram precision. Photographs were fed into four widely used apps — MyFitnessPal, Lose It!, CalAI, and Appediet. All four apps underestimated calorie content by 250–345 calories per meal on average and fat content by roughly 30 grams. According to the press release issued by the American Society for Nutrition and reported via ScienceDaily / American Society for Nutrition, the apps struggled most with ketogenic, high-fat meals — precisely the foods where fat is hidden inside the dish rather than visible on the surface.
Why the Gap Is So Large
Fat delivers more than twice the calories per gram of protein or carbohydrate, yet it is nearly invisible to a camera. Cooking oil absorbed by vegetables, butter melted into a sauce, and the marbling of a steak all disappear in a photograph. MyFitnessPal and Lose It! performed somewhat better on higher-calorie meals, while all four apps estimated carbohydrates most consistently — because bread, rice, and pasta have a predictable, photographable shape and portion size.
What This Means for Dietitians and Their Clients
A 300-calorie daily undercount accumulates to more than 2,000 extra calories per week — enough to stall or reverse weight-loss progress while the client believes they are on track. The research has not yet completed full peer review (it was presented as a conference abstract), but the controlled-kitchen design makes it among the most methodologically rigorous comparisons to date. For nutrition professionals, the practical takeaway is clear: AI photo-logging can complement dietary records, but fat-rich and mixed-dish meals still benefit from manual entry, volume estimation, or client education about the limits of camera-based tracking.