The credibility of dietary advice formulated by ChatGPT: Robo-diets for people with food allergies

Abstract

The introduction of ChatGPT has sparked enormous public interest in large language (deep-learning) models, which have been sophisticated enough to perform well on a variety of tasks. One way people are using these models is to construct diets. The prompts often include food restrictions that are an obligatory part of everyday life for millions of people worldwide. The aim of this study was to investigate the safety and accuracy of 56 diets, constructed for hypothetical individuals who are allergic to food allergens. Four levels, corresponding to the “baseline” ability of ChatGPT without prompting for specifics, as well as its ability to prepare appropriate diets when an individual has an adverse food reaction to two allergens or solicits a low-calorie diet, were defined. Findings from our study demonstrated that ChatGPT, although generally accurate, has the potential to produce harmful diets. More common errors involve inaccuracies in portions or calories of food, meals, or diets. We discuss here how the accuracy of large language models could be increased and the trade-offs involved. We propose that prompting for elimination diets can serve as one way to assess differences between such models.