Teenagers using artificial intelligence to plan their diets may be consuming dangerously few calories, equivalent to skipping a meal, a new study warns. The research, published in the journal Frontiers in Nutrition, found that AI-generated meal plans for adolescents often fall well short of recommended energy and nutrient intake.
Researchers tested five AI chatbots – ChatGPT 4, Gemini 2.5 Pro, Perplexity, Bing Chat-5GPT and Claude 4.1 – by asking them to create three-day meal plans for four 15-year-olds, including boys and girls in overweight and obese percentiles. The plans were then compared with those drawn up by a dietitian specialising in adolescent diseases.
The AI models calculated energy requirements on average nearly 700 calories lower than the dietitian did – a difference equivalent to a full meal. They also consistently disrupted the macronutrient balance, recommending around 20g more protein and 115g less carbohydrate than advised. This meant that only 32-36 per cent of energy in AI plans came from carbohydrates, against the recommended 45-50 per cent.
Lead author Ayşe Betül Bilen warned that following such unbalanced or overly restrictive meal plans during the teenage years could negatively affect growth, metabolic health and eating behaviours. She added that AI models may rely on generalised or popular diet patterns rather than fully integrating age-specific nutritional requirements.
The researchers hope the findings will raise awareness about the limited ability of AI models to develop well-balanced meal plans, particularly given that adolescence is a critical period for physical growth, bone development and cognitive maturation.
