Food Research Institute

Akduman, G. – Cebeci, A.
Guideline-based evaluation of AI-generated vs dietitian-designed meal plans in cardiovascular disease
Journal of Food and Nutrition Research, 65, 2026, No. 2, s. 133-144

Gul Akduman, Department of Nutrition and Dietetics, Faculty of Health Sciences, Burdur Mehmet Akif Ersoy University, Istiklal Campus, 15030 Burdur, Türkiye. E-mail: gakduman@mehmetakif.edu.tr

Original article
Received 24 October 2025; 1st revised 25 February 2026; accepted 1 April 2026; published online 14 April 2026

DOI: https://doi.org/10.64122/FWDF3373

Summary: Cardiovascular disease (CVD) remains a major global health burden, and diet quality is central to prevention and management. This simulation study compared artificial intelligence (AI)-generated and registered dietitian (RD)-designed one-day meal plans for 33 standardised cardiovascular risk profiles. RD plans were closer to predicted energy requirements than AI plans (median deviation –29.9 kJ vs –386.2 kJ; p = 0.017, r = 0.41). Macronutrient distribution differed. RD plans met the acceptable macronutrient distribution range for carbohydrates, protein, and fat, whereas AI plans were below the carbohydrate range and above the fat range; protein exceeded the range in 60.6 % of AI plans. AI plans also had lower plant-based protein (27.2 % vs 43.7 %) and higher saturated fat (12.4 % vs 9.5 %), along with higher n-3/n-6 fatty acids and dietary cholesterol (p ≤ 0.03), while sodium did not differ (p = 0.05). Fibre was higher in RD plans (p < 0.05). Micronutrient adequacy was generally high, but AI plans more often fell short for calcium, zinc, iron, magnesium, folate, and vitamin C. AI-generated meal plans were insufficient for cardiovascular diet therapy when used alone and should be designed or revised under RD supervision.

Keywords: artificial intelligence; cardiovascular diseases; diet therapy; dietitians; medical nutrition therapy

Download:
  (pdf, 422.47 Kb, 120x)