Stanford researchers develop AI that designs individually optimized burgers

How the blind test was conducted

The software, named BurgerAI, calculates not only taste and texture, but also the nutritional value and sustainability of the burgers – adapted to different age groups, genders, and lifestyles of the consumers. According to the researchers, the system was trained with 2,216 recipes. The team estimates the possible recipe diversity at around 10 septillion variants, i.e. a number with 43 zeros.

Particularly striking was the performance of a mushroom burger: it scored at least as well as the Big Mac in taste and surpassed it in sustainability by more than a factor of ten. A bean burger scored almost twice as high as the conventional product in terms of nutritional value. This showed that environmental friendliness and consumer acceptance do not necessarily have to be mutually exclusive.

Results for sustainability and nutritional value

First author Vahidullah Tac said: „Wir erwarteten einen Kompromiss zwischen Nachhaltigkeit und Verbraucherakzeptanz. Aber wir fanden heraus, dass ein Burger mit drastisch niedrigeren Umweltfolgen noch mit einem der weltweit erfolgreichsten Burger mithalten konnte.“ His colleague added that with such a tool one could „planetare Gesundheit und persönliche Gesundheit" zugleich treffen.

Bioengineer Ellen Kuhl, who was also involved in the study, placed the results in historical context. In the same journal, she wrote that the modern hamburger had emerged in the late 19th century from the combination of ground meat and bread. For centuries, food design had primarily been a matter of intuition, experience, and trial and error. „Jahrhunderte lang war die Gestaltung von Essen eine Sache von Intuition, Erfahrung und von Versuch und Irrtum. Wir fangen an zu zeigen, dass KI die Gestaltung von Essen in eine quantitative Wissenschaft übertragen kann", Kuhl explained.

Background: From simple ground meat and bread to AI creation

Kuhl also emphasized the complexity of the task. „Die KI hat nicht nur akzeptable Burger-Rezepte erstellt, sondern sie schuf Burger, die echte Menschen genießen", she said. This is „keineswegs trivial" and proves that the model has learned what makes foods appealing to the human palate. In the United States, an estimated 50 billion hamburgers are consumed per year – a market where the method could potentially be applied in the future.

BurgerAI as a public tool

The team around Kuhl and two other colleagues is providing the tool as a freely accessible online tool called BurgerAI. Interested users can weight parameters such as taste, nutritional value, or sustainability and have suitable recipes calculated. The researchers see this as a first step toward a more data-driven approach to food design.

Independently of this, those involved emphasized the benefit for consumers in being able to combine personal preferences with environmental aspects. If the approach proves successful in other food categories, AI could in the future also suggest individual formulations for other everyday products.