Kaya Kaçar Hüsna, Kaçar Ömer Furkan, Avery Amanda
Division of Nutrition and Dietetics, Faculty of Health Sciences, Amasya University, Amasya 05100, Türkiye.
Doctoral School of Health Sciences, Faculty or Health Sciences, University of Pécs, 7622 Pécs, Hungary.
Nutrients. 2025 Jan 7;17(2):206. doi: 10.3390/nu17020206.
With the rise of artificial intelligence (AI) in nutrition and healthcare, AI-driven chatbots are increasingly recognised as potential tools for generating personalised diet plans. This study aimed to evaluate the capabilities of three popular chatbots-Gemini, Microsoft Copilot, and ChatGPT 4.0-in designing weight-loss diet plans across varying caloric levels and genders. This comparative study assessed the diet quality of meal plans generated by the chatbots across a calorie range of 1400-1800 kcal, using identical prompts tailored to male and female profiles. The Diet Quality Index-International (DQI-I) was used to evaluate the plans across dimensions of variety, adequacy, moderation, and balance. Caloric accuracy was analysed by calculating percentage deviations from requested targets and categorising discrepancies into defined ranges. All chatbots achieved high total DQI-I scores (DQI-I > 70), demonstrating satisfactory overall diet quality. However, balance sub-scores related to macronutrient and fatty acid distributions were consistently the lowest, showing a critical limitation in AI algorithms. ChatGPT 4.0 exhibited the highest precision in caloric adherence, while Gemini showed greater variability, with over 50% of its diet plans deviating from the target by more than 20%. AI-driven chatbots show significant promise in generating nutritionally adequate and diverse weight-loss diet plans. Nevertheless, gaps in achieving optimal macronutrient and fatty acid distributions emphasise the need for algorithmic refinement. While these tools have the potential to revolutionise personalised nutrition by offering precise and inclusive dietary solutions, they should enhance rather than replace the expertise of dietetic professionals.
随着人工智能（AI）在营养与医疗保健领域的兴起，由人工智能驱动的聊天机器人越来越被视为生成个性化饮食计划的潜在工具。本研究旨在评估三款流行的聊天机器人——Gemini、Microsoft Copilot和ChatGPT 4.0——在设计不同热量水平和不同性别的减肥饮食计划方面的能力。这项比较研究使用针对男性和女性资料量身定制的相同提示，评估了聊天机器人生成的热量范围在1400 - 1800千卡的饮食计划的质量。国际饮食质量指数（DQI - I）被用于从多样性、充足性、适度性和平衡性等维度评估这些计划。通过计算与要求目标的百分比偏差并将差异分类到定义范围来分析热量准确性。所有聊天机器人的DQI - I总分都很高（DQI - I > 70），表明总体饮食质量令人满意。然而，与宏量营养素和脂肪酸分布相关的平衡子分数始终是最低的，这显示了人工智能算法存在关键局限性。ChatGPT 4.0在热量依从性方面表现出最高的精度，而Gemini的变异性更大，其超过50%的饮食计划偏离目标超过20%。由人工智能驱动的聊天机器人在生成营养充足且多样的减肥饮食计划方面显示出巨大潜力。尽管如此，在实现最佳宏量营养素和脂肪酸分布方面存在的差距强调了算法改进的必要性。虽然这些工具有可能通过提供精确且全面的饮食解决方案来彻底改变个性化营养，但它们应该增强而不是取代饮食专业人员的专业知识。
