More people are now using AI apps to keep track of their diets. All it takes is a photo of a meal, after which the app estimates its calorie and nutrient content.
However, a new study indicates that the results may be less accurate than many people believe.
Researchers tested four apps
Researchers from the National Institute of Diabetes and Digestive and Kidney Diseases tested four popular apps, according to ScienceDaily.
They compared the apps' estimates with 102 meals in which every ingredient had been weighed with great precision.
Also read: Restricted eating hours may improve cognitive function
According to the American Society for Nutrition, this provided a reliable basis for assessing how well the apps performed.
The findings were presented at the NUTRITION 2026 nutrition conference.
However, the study is still preliminary, as it has not yet been published in a peer-reviewed journal.
Certain dishes caused problems
The researchers also examined more than 200 additional meals.
Also read: A neurologist eats these five foods every day
According to the American Society for Nutrition, the apps had particular difficulty estimating dishes with a high fat content, including ketogenic meals.
By contrast, their carbohydrate estimates were more often accurate.
Underestimating dietary intake
The overall analysis showed that the four apps typically underestimated both calories and fat.
On average, the estimates were between 250 and 345 calories lower per meal, while the fat content was estimated to be around 30 grams less than the actual amount.
Also read: Why a banana may be the perfect evening snack
The researchers therefore recommend using image-based AI apps alongside more traditional methods if the goal is to track dietary intake as accurately as possible.
Also read: These foods can make your hair more fragile