AI-based calorie-tracking apps may misestimate by 345 calories

🧬 Scientific summary: Accuracy of AI-based calorie-tracking apps

A recent study showed that calorie-tracking apps that use artificial intelligence (AI) technology to analyze meal photos significantly underestimate calories and fat. According to the research findings, these apps underestimate calories by about 345 calories and about 30 grams of fat on average per meal, suggesting that their estimates may not be accurate enough to rely on alone for nutrition monitoring or weight loss.

🧠 How do AI apps estimate calories?

Apps that work with AI rely on image recognition to identify the type and quantities of food in the captured image. Once the foods and their sizes are recognized, the apps compare this data with nutrition databases to calculate calories and the rest of the nutrients.

This method provides a quick and practical alternative to manual data entry, making it popular among people who want to manage their weight or nutritional health.

🧪 Accuracy-test results for the apps using closely monitored meals

A research team conducted experiments using four popular apps: MyFitnessPal, LoseIt!, CalAI, and Appediet. These apps were tested on photos of 102 meals prepared in a clinical nutrition trial with ingredient weights measured to within 0.1 grams.

The meals in the study included one low-carbohydrate diet (the ketogenic diet) and another standard diet, providing an accurate reference to compare the estimates made by the apps with the true values.

📉 Reduced estimation accuracy in the apps

• The average calorie estimate was lower by between 250 to 345 calories per meal compared with the true values.
• Fat estimates were reduced by about 30 grams per meal.
• The apps were better at estimating carbohydrates than fats or proteins.
• Both MyFitnessPal and LoseIt! achieved better estimation accuracy for high-calorie meals than for low-calorie meals.

The results show that relying entirely on apps that depend only on analyzing a single photo without adjusting the data is misleading, especially regarding fat and the true calories consumed by the user.

🌱 Special challenges with low-carb and ketogenic meals (Keto Meals)

The initial analyses showed that the apps face greater difficulty in evaluating meals that belong to the ketogenic diet, which are characterized by a high fat content. It appears that artificial intelligence tends to underestimate the amount of fat in these meals compared with traditional meals.

As a result, the accuracy of calorie calculations in ketogenic meals becomes lower, which may affect a person’s understanding of their actual dietary intake.

🩺 The importance of integrating multiple techniques for better assessment

The research indicates that combining apps that rely on artificial intelligence for image recognition with other food-assessment methods can help improve the level of accuracy for users.

This may include manually adjusting food quantities or using more detailed methods for monitoring the diet, in order to increase the reliability of calorie and nutrient tracking.

🔍 Critical analysis and implications of the findings

This study offers one of the first approaches that smoothly compares app estimates with actual measurements in a controlled setting, which strengthens the credibility of the results.

Despite the great popularity of calorie-tracking apps, the results appear to provide underestimated estimates, especially regarding fat. This particularly affects those who rely on these apps to control weight or to ensure they consume certain amounts of nutrients.

It is important for users to know that app results are not final and should be treated critically, especially for complex meals or meals that include multiple fatty components or that rely heavily on fat, such as the ketogenic diet.

🧬 Conclusion

• AI technology in calorie tracking relies on automated image recognition to determine food content.
• Apps such as MyFitnessPal, LoseIt!, CalAI, and Appediet substantially underestimate calorie and fat estimates compared with actual measurements.
• The shortfall in fat estimation is clearly greater in meals that follow the ketogenic diet.
• Combining image-recognition tools with traditional data-entry methods improves calorie-counting accuracy.

As individuals increasingly rely on modern technology to monitor their health, this study reflects the need for further development and ongoing evaluation of AI technologies to ensure the delivery of accurate and correct information that supports informed health decisions.

📚 Related articles

🧬 What Are Statin Drugs and Why Do They Matter?

2026/09/07

A Study Shows That One Weekly Workout May Help Reduce Belly Fat

2026/09/08

FDA-Approved Drug May Enhance Immunotherapy Effectiveness Against a Rare Liver Cancer

2026/09/08