A pilot study of the potential role of ChatGPT in stated-calorie diet planning

Abstract

Developments in artificial intelligence encourage society to seek advice from artificial intelligence regarding nutrition recommendations, as in other health issues. There are not enough studies in this field. We hypothesized that ChatGPT would plan meals and daily diet within the specified calories with high accuracy. This study used ChatGPT version 3.5, freely available to the public. ChatGPT was instructed to generate daily diet plans with 1500, 2000, and 2500 calories as well as recipes with 300, 500, and 700 calories (four distinct recipe prompts were utilized for each calorie group). The calories of the resulting recipes and diet plans were calculated using nutrition databases and compared with the actual value. Only prompt-2 in the 500-calorie group showed a significant change (p < 0.05), although there was no significant difference in the four distinct recipe prompts in the 300, 500, and 700-calorie groups (p > 0.05). Among the diet plans provided by ChatGPT, there was no significant difference between the values of the 2500-calorie group and the actual calorie values in the control group (p > 0.05). According to these results, ChatGPT is more successful in creating recipe with the desired calories than daily diet planning. In this study, the calorie values of the diet plans and recipes provided by ChatGPT have demonstrated significant potential with their closeness to actual values. Further studies are needed to evaluate the reliability of ChatGPT in terms of nutritional science and the consumability of the recipes it provides.