Researchers at the University of California, Davis, have developed an AI model capable of suggesting just one to three simple ingredient substitutions to make meals healthier and more affordable without drastic changes to what people typically eat. This innovation, detailed in a study published in PLOS Digital Health, targets the difficulties many face in applying established dietary guidelines aimed at reducing disease risk.
The AI was trained on a vast dataset of over 135,000 meals from the What We Eat in America survey, identifying common meal patterns. By suggesting ingredient swaps—often including adding vegetables or legumes and replacing processed foods—the AI-generated meals aligned 47% closer to USDA nutrition targets compared to actual meals. These substitutions improved nutritional quality by about 10% and lowered meal costs by 22% to 34%.
The model outperformed existing tools, producing meals that adhered more closely to dietary recommendations while maintaining familiar flavors. Although still in the simulation phase, its potential for practical dietary improvements and public health application is promising. Researchers believe these targeted substitutions can help facilitate healthier eating without requiring complete meal redesigns, making nutritious choices more achievable for consumers.
