Clinician–Artificial Intelligence Collaboration for Mediterranean Meal-Plan Generation: Development, Technical Feasibility, and Professional Acceptability of the MAI-DIET Framework
Aug 2026· Nutrients· Vol 18, pp. 2661· 0 citations· 60 references
Medicine
TL;DR
The findings support the technical feasibility of the MAI-DIET framework and its preliminary acceptability among healthcare professionals, and further evaluation is required before conclusions can be drawn regarding its practical effectiveness.
Abstract
Background/Objectives: Adherence to the Mediterranean diet has declined in recent decades, highlighting the need for practical, technology-enabled tools to support its adoption. This study developed and evaluated MAI-DIET, a clinician-supervised, data-driven, AI-assisted programmatic framework designed to generate Greek–Mediterranean recipe-based meal plans for apparently healthy community-dwelling adults. Methods: MAI-DIET is combined with standardized food-composition and recipe libraries and a rule-based module that performs meal-plan generation and nutrient calculations. Claude Opus 4.7 supported predefined operator-supervised transformation tasks, and provided the interface for launching the rule-based module. Six 28-day recipe-based dietary plans were generated for hypothetical adults with energy goals ranging from 1600 to 2600 kcal/day. The generated plans were not tested in the intended population. For each plan, detailed nutritional analysis was performed according to predefined criteria. The quality of the dietary plans was assessed using validated scores, i.e., MedDietScore, dietary phytochemical index (DPI), and GR-UPFAST. Early-stage acceptability was evaluated by 103 healthcare professionals using a five-point Likert questionnaire. Results: All plans met the predefined energy criteria, while most nutrient targets were achieved. Deviations were observed for sodium, particularly in the higher-energy plans (reaching +36.3%), and for calcium, which was 17.5% below the EFSA population reference intake in the 1600 kcal/day plan. MedDietScore ranged between 35 and 36/55, while DPI was 47.0–50.9% and GR-UPFAST was 2.0–4.5/70. Overall acceptability was favorable (3.84 ± 0.59), with Cronbach’s alpha values of 0.838–0.952. Conclusions: The findings support the technical feasibility of the MAI-DIET framework and its preliminary acceptability among healthcare professionals. Further evaluation is required before conclusions can be drawn regarding its practical effectiveness.
BACKGROUND
Personalized meal planning by registered dietitian nutritionists (RDNs) is time-intensive. Large language models (LLMs) may automate drafting meal plans, but their nutritional accuracy in clinical practice is uncertain.
METHODS
In this proof-of-concept study, five outpatient RDNs and four LLMs (Gemini, CoPilot, ChatGPT 4.0, and customized ChatGPT 4.0) each generated 3-day meal plans for five validated clinical scenarios. Effectiveness was defined as accuracy in meeting pre-specified energy, protein, carbohydrate, fat, and sodium targets. Time to create plans and RDN comfort (self-rated confidence in nutritional accuracy and clinical appropriateness on 1-5 Likert scale) were recorded. Three independent RDNs, blinded to source, analyzed nutrient content using Nutritionist Pro. Group differences were assessed with t-test and ANOVA.
RESULTS
All LLMs and RDNs produced feasible meal plans. LLMs generated meal plans in under 1 min, whereas RDNs required a mean of 44 min per scenario. RDNs reported comfort levels ranging from 3.8 to 4.8. Across most scenarios, LLM plans delivered a smaller proportion of requested energy than RDN plans, which more consistently approached energy targets. Both groups performed similarly for the Mediterranean diet scenario. Overall, protein accuracy did not differ. However, in chronic kidney disease, LLMs undershot the guideline-based protein target, while RDNs tended to modestly exceed it. Accuracy for low-carbohydrate, fat, and sodium diets was comparable.
CONCLUSION
LLMs can rapidly generate clinically plausible meal plans but are less reliable than RDNs in achieving prescribed energy and selected macronutrient goals. Prompt precision is essential for nutrient-specific targets. A hybrid model in which RDNs refine LLM-generated drafts may leverage efficiency without sacrificing clinical accuracy.
M. Mundi, Osman Mohamed Elfadil, Danielle P. Johnson et al.· Nutrition in clinical practi...· 0 citations
This paper presents the development of a diet plan generator system that uses user-specific parameters such as age, weight, gender, activity level, dietary goals, and food preferences to recommend a structured meal plan.
Translating nutritional recommendations into practical day-to-day meal choices remains a challenging task, particularly when personalization, nutritional adequacy, dietary diversity, allergies, seasonal availability, and food-group constraints must be simultaneously satisfied. This study presents and evaluates the PLAN'EAT Nutrition Advisor, an Artificial Intelligence (AI)-driven, expert rule-based nutrition recommendation system, designed to generate personalized and nutritionally balanced weekly meal plans aligned with established dietary guidelines and food-group recommendations derived from Sustainable Healthy Diet (SHD) principles. The proposed approach is built upon the PLAN'EAT Expert-Curated Meal Database, a nutritionist-designed repository introduced in this work, comprising 401 expert-curated meals spanning Irish, Spanish, and Hungarian cuisines. Meal-plan generation follows a four-stage pipeline: (1) meal filtering based on country-specific cuisine, seasonality, dietary preferences, and allergies, (2) daily meal plan generation through large-scale sampling and scoring against expert-defined nutritional targets, (3) weekly meal plan assembly and optimization under nutritional and food-group constraints, and (4) diversity optimization to promote dietary variety while preserving nutritional validity. The proposed approach was validated through a large-scale in-silico validation involving 1,000 synthetic user profiles and the generation of 16,000 weekly meal plans corresponding to 112,000 daily meal plans. Adherence to daily and weekly nutritional targets, food-group constraints, and overall meal-plan validity at scale were assessed. In addition, we performed a preliminary descriptive evaluation against state-of-the-art Large Language Model (LLM)-based approaches under two complementary settings: Retrieval-Augmented Generation (RAG) and Supervised Fine-Tuning (SFT). Experimental results demonstrate that the proposed system can efficiently generate personalized, nutritionally compliant, and diverse weekly meal plans while maintaining transparent expert-rule-driven optimization. Furthermore, the descriptive evaluation against LLM approaches suggests that the proposed system achieves more consistent energy and macronutrient adherence under the evaluated conditions.
Dimitris Tsolakidis, Vasilis Stamatis, L. Gymnopoulos et al.· Frontiers in Nutrition· 0 citations
Artificial intelligence (AI)-based large language models (LLMs) are increasingly used to support nutrition-related decision-making; however, their ability to generate clinically appropriate dietary plans for inherited protein metabolism disorders remains largely unexplored. This study aimed to perform an exploratory simulation-based benchmarking analysis of AI-generated dietary plans for phenylketonuria (PKU), maple syrup urine disease (MSUD), and propionic acidemia (PPA) using disease-specific metabolic nutrition guidelines.
Standardized pediatric case scenarios were developed for PKU, MSUD, and PPA. Using identical English-language prompts, ChatGPT-5.3 Pro and Gemini 3 Pro Advanced each generated 3-day dietary plans. Nutrient composition was analyzed using the BeBiS Nutrition Information System and evaluated against Dietary Reference Intakes (DRIs). Disease-specific nutritional targets, including amino acid intake, protein distribution, and energy provision, were benchmarked against recommendations from Genetic Metabolic Dietitians International (GMDI). Nutritional characteristics of the dietary plans generated by the two AI models were compared using exploratory statistical analyses.
Both LLMs generated structured dietary plans with generally acceptable overall nutritional characteristics; however, clinically relevant deviations from disease-specific nutritional targets were identified across all three disorders. In the PKU case, both models achieved the recommended phenylalanine range, but neither simultaneously met protein and tyrosine recommendations. In the MSUD case, differences were primarily related to energy provision and branched-chain amino acid targets, while in the PPA case neither model achieved the recommended balance between intact protein and total protein. These findings demonstrated that conventional measures of nutritional adequacy alone were insufficient to determine the clinical appropriateness of AI-generated dietary plans for inherited protein metabolism disorders.
General-purpose LLMs can generate structured dietary plans for inherited protein metabolism disorders; however, disease-specific metabolic targets are not consistently achieved. Evaluation of AI-generated dietary plans should therefore extend beyond conventional nutritional assessment and incorporate disease-specific benchmarking against established metabolic nutrition guidelines. This study provides a disease-specific benchmarking framework for evaluating AI-generated dietary plans in inherited protein metabolism disorders.
Taha Gökmen Ülger, E. Adıgüzel· Frontiers in Nutrition· 0 citations
INTRODUCTION
Pharmacotherapy optimization in multimorbid patients is increasingly complex due to polypharmacy, fragmented data, expanding electronic health records, and workforce constraints. Conventional clinical decision support systems remain largely rule-based and often fail to adequately incorporate patient-specific context. While artificial intelligence offers new opportunities, stand-alone models remain insufficiently reliable for high-risk pharmacotherapy decision support.
AIM
To develop a relevance-driven, clinician-supervised hybrid AI framework for pharmacotherapy optimization.
METHOD
Using a design science-informed approach, an interdisciplinary research group developed a conceptual framework for AI-supported pharmacotherapy optimization. Framework development was informed by prior feasibility work, published literature, clinical practice requirements, and iterative interdisciplinary discussions. Hybrid AI was defined as the combination of retrieval-augmented generation, deterministic safety rules, and large language model reasoning.
RESULTS
Seven design principles were identified, including decomposition of clinical activities, relevance-based prioritization, hybrid reasoning under clinician oversight, integration of patient goals, transparency of evidence sources, longitudinal optimization within a governed closed loop, and evaluation as a design requirement. These principles informed a conceptual architecture integrating structured clinical data, patient preferences, longitudinal patient information, and evidence retrieval within a clinician-governed decision-support framework.
CONCLUSION
The proposed framework conceptualizes AI as a relevance-structuring, clinician-governed decision-support layer rather than an autonomous decision-maker. By combining hybrid reasoning, patient-specific context, and professional oversight, it provides a conceptual foundation for future development, implementation, and evaluation of AI-supported pharmacotherapy systems.
Olaf Rose, Stephanie Clemens, A. Leiherer et al.· International Journal of Cli...· 0 citations
Artificial intelligence is drastically entering every aspect of our lives, including education, training, and communication. There are many models in Generative artificial intelligence (GenAI) such as “large language models (LLMs) including ChatGPT, Gemini, Claude, and Copilot etc. This narrative review focuses on the current peer-reviewed articles (predominantly 2023–2026) on the AI applications, opportunities, challenges and future directions in the field of nutrition education. Most of these applications span three domains: professional training, direct-to-consumer dietary counselling and meal planning (chatbot-delivered advice, personalised diet plans), and public health nutrition communication. Evidence suggests that GenAI can improve nutrition learning's accessibility, scalability, personalisation, and engagement while providing dietetics educators with an affordable substitute for resource-intensive training techniques like standardised patients. However, studies consistently show limitations in accuracy, especially when it comes to calculations of calories and macronutrients, complex or comorbid clinical scenarios, and culturally specific dietary contexts. These limitations are accompanied by concerns about misleading information, algorithmic bias, dependency, data privacy, and the deterioration of critical thinking. Future directions include multimodal food-image analysis, retrieval-augmented generation based on validated nutrition databases, hybrid human-AI counselling models, and formal AI-literacy curriculum for the general public and dietetics students. According to the review, GenAI works best as a scalable supplement to trained nutritionists rather than as a replacement for them.
Sameeksha Sharma· International Journal of Nut...· 0 citations