Ayşe Betül Bilen, PhD, notes metabolic and developmental risks of AI-generated diets

In a recent interview with Contemporary Pediatrics, Ayşe Betül Bilen, PhD, assistant professor at İstanbul Atlas University, breaks down the findings of a new study evaluating the safety of artificial intelligence (AI)–generated diet plans for adolescents.

With teens increasingly turning to tools such as ChatGPT and Gemini for fitness and nutrition advice, the research reveals that these AI models consistently produce severely imbalanced, low-carbohydrate diets while undercalculating daily energy needs by an average of 700 kilocalories. Bilen highlights the immediate developmental risks of these sustained deficits—including compromised bone density, stunted pubertal growth, and impaired cognitive function—and emphasizes the need for pediatricians and dietitians to proactively guide families toward evidence-based nutrition rather than relying on flawed, unsupervised AI recommendations.

Contemporary Pediatrics:

The study revealed that AI models skewed macronutrient ratios significantly, recommending higher protein and lipid ratios while severely dropping carbohydrates. What are the most immediate metabolic and developmental risks for an adolescent who attempts to follow this type of unsupervised, low-carb/high-fat diet during their peak growth periods?

Ayşe Betül Bilen, PhD:

Adolescence is a critical period of rapid growth, hormonal maturation, and brain development. Diets that are too low in carbohydrates and relatively high in fat and protein can disrupt the balance of nutrients needed to support these processes. Carbohydrates are an important energy source for the brain and for daily activity. When adolescents follow overly restrictive or imbalanced diets without professional guidance, it may affect energy levels, growth patterns, and healthy eating behaviors.

Contemporary Pediatrics:

With the AI tools systematically undercalculating energy needs by an average of nearly 700 kcal, how might this substantial and sustained caloric deficit impact pubertal development, bone density acquisition, and cognitive function in adolescents if they rely solely on these tools?

Ayşe Betül Bilen, PhD:

In our study, AI-generated plans underestimated adolescents’ daily energy needs by about 700 kilocalories on average. If such a deficit were sustained over time, it could interfere with normal growth and pubertal development. Adolescence is also the period when a large proportion of lifetime bone mass is formed. Inadequate energy intake during this stage may negatively affect bone health and could impact concentration, cognitive performance, and overall well-being.

Contemporary Pediatrics:

The findings note significant variations and inconsistencies in micronutrient content across all models. Given the critical importance of micronutrients such as iron, calcium, and vitamin D during adolescence, which specific deficiencies worry you the most when teens turn to flawed AI generators rather than clinical guidelines?

Ayşe Betül Bilen, PhD:

Several micronutrients are particularly important during adolescence, including iron, calcium, vitamin D, and zinc. These nutrients play key roles in bone development, oxygen transport, immune function, and overall growth. When diet plans are inconsistent or poorly balanced—as we observed in several AI-generated plans—there is a risk that adolescents may not meet these critical nutrient requirements.

Contemporary Pediatrics:

Since adolescents and their parents increasingly use accessible AI models such as ChatGPT or Gemini for quick fitness or diet plans, how should pediatricians and registered dietitians proactively address the AI diet trend and its clinically significant deviations during routine wellness visits?

Ayşe Betül Bilen, PhD:

Health professionals should acknowledge that many adolescents already use digital tools for health information. Instead of dismissing AI entirely, pediatricians and dietitians can use this as an opportunity to educate families about evidence-based nutrition. During routine visits, clinicians can discuss how AI-generated information should be interpreted cautiously and emphasize the importance of individualized guidance for adolescents.

Contemporary Pediatrics:

The conclusion emphasizes that AI dietary recommendations are unsafe without professional supervision. Looking forward, what specific guardrails, data training improvements, or clinical integrations need to happen before we can safely use AI as an assistive tool in pediatric nutrition?

Ayşe Betül Bilen, PhD:

AI has the potential to support nutrition education and provide accessible information, but it needs stronger safeguards before being used for clinical decision-making. Future improvements should include better integration of evidence-based dietary guidelines, age-specific nutritional requirements, and clearer warnings when advice may not be appropriate for minors. Ideally, AI tools should function as supportive educational resources rather than replacements for professional dietary counseling.

No relevant disclosures.

Reference

Bilen AB, Kalkan GE, Önal HY. Artificial intelligence diet plans underestimate nutrient intake compared to dietitians in adolescents. Front Nutr. 2026;13. doi:10.3389/fnut.2026.1765598