Calorie tracking apps powered by artificial intelligence can estimate the nutritional content of a meal from a single photo. The technology offers a quick and convenient alternative to manually introducing each food and serving, but new research suggests the results may be significantly inferior to what’s actually on the plate.
In a test of four photo-based apps, researchers found that calorie and fat estimates were about a third too low on average.
How AI estimates calories from food photos
Photo-based calorie tracking relies on AI image recognition to identify foods shown in an image and estimate the size of each serving. The app then compares those estimates to nutritional databases to calculate calories and other nutrients.
“Photo-based calorie tracking apps are very popular, especially for people trying to monitor their health or lose weight,” said Aaron Hengist, a visiting postdoctoral fellow in the Intramural Program at the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), part of the National Institutes of Health. “However, the accuracy of many of these apps has not been thoroughly evaluated. Our study helps address this question by looking at whether these apps can reliably estimate calories.”
Olivia Charles, a graduate intramural research training fellow at NIDDK, presented the findings at NUTRITION 2026, the flagship annual meeting of the American Society for Nutrition, held July 25-28 in National Harbor, Maryland, just outside Washington, DC.
Test apps with precisely measured meals
The project is part of a larger nutrition study at the NIH Clinical Center that is investigating how the body processes nutrients on either a low-carbohydrate (ketogenic) diet or a standard diet.
The meals used in the clinical trial are prepared in a controlled metabolic kitchen, where researchers measure ingredients to the nearest 0.1 gram. This gave the team a very precise baseline for evaluating applications.
The researchers collected standardized photographs of 102 prepared meals for the diet study. They then sent the images to MyFitnessPal, LoseIt!, CalAI and Appediet to see how well each app’s estimates matched the known nutritional content.
“By using meals prepared in a tightly controlled metabolic kitchen, we were able to compare the apps’ estimates with an accurate reference,” Hengist said. “This type of high-quality, direct comparison has not been available before.”
Apps lost hundreds of calories
Across all four apps, the estimated total calories were between 250 and 345 calories too low per meal on average. The apps also underestimated fat by about 30 grams.
MyFitnessPal and LoseIt! They were more accurate when analyzing high-calorie foods than when analyzing low-calorie foods. All four applications also produced more consistent estimates for carbohydrates than for other macronutrients.
“People who use a photo-based tracking app without adjusting portions or entering food quantities should take the results with a grain of salt,” Hengist said. “These apps tend to underestimate calories, especially from fat, so what they actually ate is probably higher than what the app shows.”
Ketogenic Meals May Be Harder for AI to Measure
After the first analysis, the researchers tested more than 200 additional meals to investigate what factors could influence the accuracy of the application.
Preliminary findings indicate that apps may have greater difficulty evaluating low-carb ketogenic diet meals. These meals tend to contain more fat, which the apps tended to consistently underestimate.
The researchers suggest that combining photo-based tools with traditional methods for assessing food intake and diet quality could make calorie tracking more accurate in everyday use.
Charles presented this research on Saturday, July 25 during the President’s Oral Session at the Gaylord National Resort & Convention Center (abstract).
The abstracts presented at NUTRITION 2026 were reviewed and selected by a committee of experts. However, they generally have not completed the full peer review process required for publication in a scientific journal. Therefore, the results should be considered preliminary until they appear in a peer-reviewed publication.