ChatGPT-4o for weight management: Comparison of different diet models

Abstract

In recent years, artificial intelligence (AI) tools such as ChatGPT have emerged as accessible and scalable platforms for generating dietary advice. While ChatGPT has demonstrated potential in providing general nutritional guidance, its capacity to create diet plans tailored to different weight categories and physical activity levels remains underexplored, particularly in comparison across popular dietary (ketogenic and intermittent fasting) models. This study aimed to evaluate the nutritional adequacy and variability of diet plans generated by ChatGPT-4o for weight management. ChatGPT-4o generated diet plans for 18 individuals (9 males, 9 females) representing overweight, class I, and class II obesity at varying physical activity levels. Fifty-four menus were created across three dietary models and analyzed for energy, macro-, and micronutrient content using the BeBiS nutritional analysis software. Diet variability was also assessed through repeated prompts over three different periods. The ketogenic diets produced by AI had significantly higher energy and saturated fat contents than other models (p < 0.05). Regardless of prompt, AI often produced low-carbohydrate, high-fat diets. The menus created by ChatGPT had significantly higher fat, saturated fat, and protein content but lower carbohydrate content compared to the dietitian menus (p < 0.05). Micronutrient analysis showed frequent inadequacy in calcium, potassium, and vitamin B1. Notably, menu content showed temporal inconsistencies, particularly in intermittent fasting and ketogenic diets. While ChatGPT-4o shows promise in generating basic dietary models, concerns remain about its nutritional precision, consistency, and safety. The results revealed the necessity of professional supervision in AI-assisted nutrition planning.