AI chatbots are giving teenagers dangerous advice about eating and producing diet plans that don’t provide enough nutrients or calories, a new study has claimed.
Researchers from Istanbul Atlas University told AI models, including ChatGPT and Gemini, to create meal plans for teenagers trying to lose weight and found that, on average, they provided diets with almost 700 calories less than that of a dietitian.
They are now warning teenagers against using AI tools to create diet plans and have called for the development of safer tools with the help of professionals.
Dr Ayşe Betül Bilen, the study’s author, said: “We show that diet plans generated by AI models tend to substantially underestimate total energy and key nutrient intake when compared to guideline-based plans prepared by a dietitian.
“Following such unbalanced or overly restrictive meal plans during the teenage years may negatively affect growth, metabolic health, and eating behaviours.”
The researchers looked at meal plans provided by ChatGPT 4, Gemini 2.5 Pro, Bing Chat-5GPT, Claude 4.1, and Perplexity and compared them against the recommendations of a registered dietician.
The Independent has contacted the AI companies mentioned in the study for comment.
Researchers instructed the chatbots to create meal plans based on a person’s specific age, height, and weight. They told them to create a plan for three days, covering three meals and two snacks per day.
Meal plans were created for four 15-year-olds, including a boy and a girl falling into the overweight percentile, and a boy and a girl who were falling into the obese percentile.
On average, the meal plans made by AI had 700 calories less than those made by a dietician. Researchers said the difference is equivalent to a full meal and could have serious clinical consequences.
AI also recommended a significantly lower carbohydrate intake of between 32 and 36 per cent. Many nutritionists recommend that around 50 per cent of someone’s energy intake come from carbs.
They also found the chatbots suggested a protein intake that was around 20g higher than that of the dieticians.
Dr Bilen said: “Adolescence is a critical period for physical growth, bone development, and cognitive maturation. Lower energy and carbohydrate intake, combined with increased protein and fat ratios, may pose risks during the adolescent growth period.
“AI models are primarily trained to generate responses that appear plausible and user-friendly rather than clinically precise. Our findings suggest they may rely on generalized or popular diet patterns instead of fully integrating age-specific nutritional requirements,” he added.
Anastasia Kalea, a senior nutrition researcher from University College London, said: “While digital platforms provide a stigma-free sanctuary for people living with obesity, the current reliance on AI tools presents a significant risk: many of these models depend on generalised, publicly available data rather than the nuanced diet and medical history information of the individual which would be collected meticulously by a healthcare professional trained in dietetics and nutrition.
“This lack of personalisation in such scenarios makes treatments less effective and harder to maintain for long term compared to the care of a trained dietitian, who investigates the root causes and socioeconomic layers of this complex disease through active listening and clinical judgment. The problem though is not lack of efficiency, as much as the risk of causing harm.”
