THURSDAY, March 12, 2026 (HealthDay News) -- For adolescents, artificial intelligence (AI) model-generated diets exhibit clinically significant deviations from dietitian reference plans, according to a study published online March 12 in Frontiers in Nutrition.
Ayşe Betül Bilen, from Istanbul Atlas University, and colleagues evaluated the clinical validity of AI by comparing the nutritional content of diets generated by different AI models with dietitian reference plans. In the cross-sectional and comparative study, 60 three-day diet plans were generated in two sessions by five AI models for four standardized adolescent profiles. For each profile, a dietitian reference plan was prepared.
The researchers found that AI models tended to systematically undercalculate energy, protein, lipid, and carbohydrate (mean bias, +695 kcal, +19.9 g, +15.8 g, and +114.6 g, respectively). In macronutrient percentages, protein and lipid ratios were above (21.5 to 23.7 percent and 41.5 to 44.5 percent, respectively) and carbohydrate ratios were below (32.4 to 36.3 percent) the recommended adolescent guidelines. Micronutrient contents varied significantly between models; none of the models showed consistent proximity to the dietitian for any nutrients.
"Adolescence is a critical period for physical growth, bone development, and cognitive maturation," Bilen said in a statement. "Lower energy and carbohydrate intake, combined with increased protein and fat ratios, may pose risks during the adolescent growth period."