Background Effective dietary management is essential for individuals with type 1 diabetes (T1D). Artificial intelligence (AI) tools such as ChatGPT-4o, Bard AI, and Bing AI are increasingly being used to assist in healthcare tasks, including nutrition advice. This study evaluates the performance of these AI models in generating dietary recommendations when compared to input from human dietitians. Methods Sixty expert-written, hypothetical T1D patient cases were submitted to ChatGPT-4o, Bard AI, and Bing AI. Each model’s responses were assessed as either “Correct” or “Incomplete” relative to dietitian recommendations. Descriptive statistics and McNemar’s test were used to compare pairwise performance across models. Results ChatGPT-4o provided correct recommendations in 60% of cases, followed by Bard AI (50%) and Bing AI (26. 7%). McNemar’s test showed that ChatGPT-4o significantly outperformed Bing AI (p 0. 05). ChatGPT-4o demonstrated superior resilience across case complexity levels, showed the highest rate of unique correct answers, and exhibited only modest agreement with other models. This highlights ChatGPT-4o’s relative independence and robustness. An interactive version of the analysis can be accessed here: https: //cnpdata. shinyapps. io/aidiabetes/ Conclusions ChatGPT-4o generated more accurate dietary suggestions than Bing AI and performed comparably to Bard AI. However, AI tools still lack the contextual nuance of human dietitians and should be used to supplement, rather than replace, professional guidance in diabetes care.
Nasser Alqahtani (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: