New Research Warns AI-Generated Diet Plans Consistently Fail to Meet Critical Nutritional Requirements for Growing Adolescents

Research shows AI tools like ChatGPT underestimate teen caloric needs by 700 kcal, often pushing imbalanced, low-carb diets that risk developmental harm.

By: AXL Media

Published: Mar 20, 2026, 11:27 AM EDT

Source: Information for this report was sourced from Dr. Liji Thomas, MD

The Growing Disconnect Between AI Advice and Pediatric Nutrition

As adolescent obesity rates climb globally, many young people are turning to artificial intelligence for accessible weight management advice. However, a new investigation conducted by medical experts indicates that these digital tools are not yet equipped to handle the complex nutritional needs of the teenage population. The study compared 60 diet plans generated by five major AI models against professional dietitian standards, uncovering a systematic failure to align with international health guidelines. While AI offers the allure of instant, personalized planning, the data suggests it consistently falls short of the caloric thresholds required to sustain healthy growth during puberty.

Significant Shortfalls in Caloric and Macronutrient Intake

The most alarming finding of the research was a consistent "energy gap" in the AI-generated recommendations. On average, the AI models provided nearly 700 kcal less per day than the reference plans designed by human dietitians. Furthermore, the protein and fat distributions were significantly skewed, with AI models underestimating carbohydrate needs by approximately 115 grams. This shortfall is particularly concerning for adolescents, whose high metabolic demands support not only physical growth but also critical brain development. Medical professionals warn that such a substantial deficit in energy could lead to long-term metabolic and developmental setbacks.

Concerns Over Keto-Leaning Bias in Large Language Models

Researchers observed that the AI models appeared to rely more on popular "fad" diets, such as ketogenic or high-protein regimes, rather than established scientific guidelines for children. The AI-generated plans frequently pushed fat content up to 44.5%, while suppressing carbohydrates to as low as 36.3%. In contrast, professional dietitian plans maintained a balanced carbohydrate range of 44% to 46% to meet national health recommendations. This systematic shift toward low-carb, high-lipid meal structures suggests that AI algorithms may be "over-learning" from internet-based diet culture rather than prioritizing the specific physiological requirements of a developing body.