AI-generated diets for adolescents can be overly restrictive and nutritionally unbalanced, with the real risk of encouraging unhealthy eating behaviors. This is the key finding of a study published in Frontiers in Nutrition, which compared meal plans generated by different AI systems with those developed by a dietitian. The results indicate a systematic tendency to excessively reduce calories and to favor proteins and fats at the expense of carbohydrates. This approach, especially during a delicate stage like adolescence, can have implications not only on a metabolic level but also on a psychological level.
A Comparative Study Between AI And Dieticians
The work, conducted by Ayşe Betül Bilen and colleagues at Istanbul Atlas University, represents one of the first systematic analyses directly comparing meal plans generated by different AI models with those developed by a dietitian for overweight or obese adolescents.
The researchers asked five AI models to create three-day meal plans for four profiles of fifteen-year-olds, using standard parameters such as age, weight, and height. The same scenarios were then run by a clinical nutrition specialist, to obtain a direct and methodologically sound comparison. The result was clear: the AI systems failed to replicate the accuracy and balance of the plans constructed by a professional.
Too Much Protein and Fat, Too Few Carbohydrates and Calories
The most striking finding concerns energy requirements: AI-generated diets systematically underestimate calorie needs. On average, they "allow" approximately 700 kcal fewer than recommended by a dietitian. A difference of this magnitude is not insignificant: it's effectively equivalent to cutting out a full meal.
In adolescence – a stage characterized by rapid growth, hormonal development, and brain maturation – a prolonged caloric deficit can have significant metabolic and functional consequences.
Beyond quantity, a qualitative problem emerges. AI models tend to propose diets unbalanced in macronutrients: a protein content higher than recommended (over 21% of total energy), a very high fat intake (up to 44–45%), and a significant reduction in carbohydrates (around 32–36%).
This pattern differs significantly from developmental guidelines, which require a balance capable of supporting both energy needs and cognitive and physiological functions. In particular, restricting carbohydrates can compromise energy availability for the brain and daily activities.
Health Risks and Eating Habits
The implications aren't limited to immediate nutrition: according to the authors, following overly restrictive or unbalanced eating plans over time can negatively impact adolescents' growth, metabolic health, and even eating behavior.
In this age group, already exposed to body image pressures, the unsupervised use of automatic devices can become a risk factor for the development of eating disorders. The combination of calorie deficits and rigid dietary patterns can foster a dysfunctional relationship with food.
The key issue that emerged from the study concerns the very nature of artificial intelligence: algorithms are capable of processing numbers and proportions, but lack a true clinical understanding of the individual. An effective nutritional plan, especially during adolescence, requires an integrated assessment of many variables: medical history, pubertal development, family context, daily habits, and psychological aspects. All elements that, at present, cannot be adequately modeled by an automated system.
The study does not suggest abandoning artificial intelligence in the nutritional field, but invites us to reduce its role: digital tools can be useful for information and support, but they should not replace specialist assessment, especially during developmental age.
The final message is clear: diets are not simple mathematical schemes, but complex therapeutic interventions. Entrusting them to unsupervised systems, during a delicate stage like adolescence, can transform a technological opportunity into a concrete health risk.
;Resize,width=767;)
