Clinical Study Warns AI-Generated Diet Plans Provide Inadequate Nutrition and Imbalanced Macronutrients for Developing Adolescents

New research shows AI tools like ChatGPT underestimate teen caloric needs by 700 kcal, favoring imbalanced high-fat diets over expert medical guidelines.

By: AXL Media

Published: Mar 23, 2026, 6:58 AM EDT

Source: Information for this report was sourced from [Dr. Liji Thomas, MD / Frontiers in Nutrition]

Clinical Study Warns AI-Generated Diet Plans Provide Inadequate Nutrition and Imbalanced Macronutrients for Developing Adolescents - article image
Clinical Study Warns AI-Generated Diet Plans Provide Inadequate Nutrition and Imbalanced Macronutrients for Developing Adolescents - article image

Systemic Underestimation of Caloric Requirements

Artificial intelligence models are failing to account for the high energy demands essential for adolescent development. A comparative study of five major AI tools—ChatGPT-4o, Gemini 2.5 Pro, Claude 4.1, Bing Chat-5GPT, and Perplexity—found that these models generated diet plans with an average energy shortfall of 695 kcal compared to professional dietitian standards. This significant caloric gap suggests that AI tools may inadvertently promote under-nutrition in a demographic that requires substantial energy for physical growth and metabolic stability.

Distortion of Macronutrient Ratios and Health Guidelines

The research highlights a "systematic shift" in AI-generated nutrition where macronutrient balances are significantly disrupted. While professional dietitians align plans with international guidelines—typically 45–50% carbohydrates—AI models frequently recommended diets where carbohydrates accounted for as little as 36.3%. Conversely, protein and fat content in AI plans often exceeded recommended levels, reaching up to 23.7% and 44.5% respectively. This trend indicates that Large Language Models (LLMs) may be over-relying on popular "fad" diets, such as ketogenic patterns, rather than evidence-based pediatric nutrition science.

Micronutrient Variability and Clinical Safety Risks

Beyond basic calories, the micronutrient composition of AI-generated diets showed extreme volatility and failed to meet safety standards. No AI model tested was able to closely adhere to a dietitian’s reference diet across all essential vitamins and minerals. Dr. Liji Thomas notes that these inadequacies could hinder growth, metabolism, and even cognitive development during the crucial adolescent window. Because these technologies lack technical expertise in pediatric clinical nutrition, their unsupervised use for weight management in overweight or obese youth is currently deemed unsuitable by researchers.

Categories

Topics

Related Coverage