The integration of artificial intelligence-generated diets into the field of dietetics: Synergy or threat?

The objective of this study was to conduct a comparative analysis of diets generated by artificial intelligence (AI) tools (ChatGPT-4o and Gemini Advanced) and those prescribed by dietitians. Each researcher (n: 3) obtained 15-day diet plans from AI in separate accounts and 3-day diet plans from five dietitians. In total, 135 diet plans were obtained for each researcher, 45 from ChatGPT-4o, 45 from Gemini, and 45 from dietitians. Diet plans were entered into the BeBis v9.0 program. Nutrient Rich Foods (NRF)15.3 scores, estimated glycemic index(GI) score, saturated fatty acid(SFA)/unsaturated fatty acid ratio, MUFA/PUFA ratio, ω-6/ω-3 ratio, amino acid (AA) score, and composite dietary antioxidant index(CDAI) score were evaluated. ChatGPT-4o achieved the highest trient NRF15.3 score, while Gemini Advanced achieved the highest AA score despite overall inadequate protein quality. AI tools exhibited a significantly lower percentage of energy from SFA(%), SFA/USFA, and SFA/PUFA ratio in comparison with diets prescribed by dietitians. The diets produced by Gemini Advanced demonstrated the highest MUFA/PUFA ratio. All diet plans were within the acceptable range in terms of the ω-6/ω-3 ratio, with no significant differences between the groups. While no significant differences were found in the estimated GI, dietitian-prescribed diets achieved significantly higher CDAI, suggesting greater antioxidant capacity in their overall menu composition. These findings suggest that AI tools can generate nutritionally sound diets with only minor deviations from those prepared by dietitians, potentially enhancing efficiency and reducing workload; however, human oversight remains essential to ensure nutritional adequacy, address micronutrient gaps, and guide critical dietary decisions.