A generative AI model trained on more than 2,200 burger recipes has produced new formulations that optimize taste, nutrition and environmental impact, according to a study published in npj Science of Food. The research demonstrates how AI can help developers navigate trade-offs that have long stalled plant-based product innovation, identifying ingredient combinations that improve sustainability without sacrificing flavor.
The model differs from large language models that output recipe text. The researchers built a diffusion-based AI model that learned ingredient combinations and quantities from existing burger recipes. It then generated one million new formulations, each screened for palatability, nutritional quality and environmental performance.
Balancing taste and sustainability
Among the recipes chosen for testing, a mushroom-based burger achieved an environmental impact score more than ten times lower than a conventional Big Mac reference. A mushroom and beef hybrid performed similarly to the Big Mac in blind taste tests while reducing environmental impact, indicating that strategic ingredient blends-not one-for-one substitutions-can deliver sustainability gains.
The model also produced a bean-based burger with nearly double the Healthy Eating Index score of the Big Mac and roughly one-sixth of its environmental impact. However, participants rated it lower for taste and texture, underscoring that consumer experience remains the central challenge.
Consumer acceptance remains the hurdle
To test whether AI-generated recipes translated into palatable products, researchers ran a blind sensory trial with 101 participants in San Francisco. Two recipes developed to maximize palatability received flavor or overall liking scores that matched or exceeded those of the Big Mac, while texture ratings were comparable. These results suggest AI can help pinpoint formulations that improve acceptance-a persistent barrier for plant-based foods.
Personalized recipes for targeted nutrition
The research also showed that the model could generate personalized burger recipes based on age, sex and activity level. This capability points to broader commercial uses, where product developers might tailor foods to specific demographic or nutritional profiles without starting from scratch.
Why this matters for product development professionals
For teams working in AI for Product Development, the study offers a blueprint for using generative models to accelerate formulation R&D. Instead of iterating through costly physical prototypes, developers can simulate thousands of ingredient combinations and screen for multiple objectives simultaneously. The ability to incorporate sensory and environmental criteria early in the design process could reduce time to market and increase the odds of launching products that consumers embrace.
Your membership also unlocks:
