Your AI calorie-tracking app may be off by 345 calories

AI-powered calorie tracking apps are becoming a common tool for health management, but new research suggests they lack precision. A study presented at NUTRITION 2026 found that four popular photo-based apps consistently underestimated calorie and fat totals by nearly one-third when tested against controlled, precisely measured meals.

Researchers from the National Institute of Diabetes and Digestive and Kidney Diseases evaluated apps including MyFitnessPal, LoseIt, CalAI, and Appediet. By comparing app estimates against meals prepared in a metabolic kitchen, the team found that these tools missed an average of 250 to 345 calories per meal. The data indicates that fat content, particularly in high-fat or ketogenic dishes, presents a significant hurdle for current image recognition technology. While carbohydrate measurements proved more stable, the broad margin of error remains a concern for users counting calories for specific dietary goals.

The findings highlight a gap between the convenience of quick photo uploads and the reality of nutritional accuracy. Although apps simplify the process of logging meals, they frequently misjudge portion sizes or ingredient densities. These discrepancies are often large enough to impact weight management progress, as a 300-calorie daily deficit or surplus adds up significantly over time.

Experts recommend that users view these AI-generated estimates as rough guides rather than precise data. Combining these apps with traditional tracking methods, or manually verifying portion sizes, remains a necessary step for those who require accurate caloric intake logs. Relying solely on a single photograph to dictate a daily nutrition plan is risky given the current limitations of the software.