Assessing the quality of diet and exercise plans generated by AI Chatbots: a preliminary study using the NExGEN prompt generator system / Azwa Suraya Mohd Dan ... [et al.].

Artificial intelligence (AI) chatbots like ChatGPT are increasingly used in obesity research to track diets, activity, and energy expenditure. However, its effectiveness in diet and exercise planning depends on the precision and completeness of user inputs. This study evaluates the quality of ChatGPT output when combined with the newly developed diet and exercise prompt generator system, NExGEN. A cohort of obese participants (n = 18) was enlisted to contribute interpersonal data for the NExGEN prompt generator. Utilising ChatGPT-4, this data informed the creation of 36he36ious36s36d weekly dietary and exercise plans. Accredited professionals (n = 16) conducted a blind evaluation of these plans by grading the quality and validity of the NExGEN-ChatGPT responses using the DISCERN and content validity index (CVI), respectively. The evaluators graded the NExGEN-ChatGPT responses as bottom tier 2.2% of the time, middle tier 16.3% of the time, and top tier 81.5% of the time. The CVIs score was ≥ 80% with a correlation coefficient between 0.89 – 0.99, and overall Cronbach’s alpha score at 0.798. This study demonstrates that integrating ChatGPT with the NExGEN system effectively generates high-quality diet and exercise plans for obese individuals, as evidenced by favourable quality and validity assessments by professionals.