In a San Francisco restaurant, a hundred and one people quietly compared five burgers designed by a generative AI model against the most familiar burger of the modern age, the Big Mac, not knowing which was which. The effort, reported in npj Science of Food by researchers at Stanford, belongs to an old human struggle: to eat in ways that feed people well without wounding the land and water that sustain them. Two machine-made burgers edged past the classic in flavor and liking, though only by modest margins, while the designs chasing climate or health goals tended to give up some of the pleasur
AI-Designed Burgers Edge Out the Big Mac on Taste in Blind Test, but Trade-Offs Remain
No single burger won on taste, nutrition, and sustainability together.
So the headline is that AI beat the Big Mac. How big was the win?
Small. Two burgers scored slightly higher on flavor, and one also on overall liking. The differences were statistically significant but modest, and texture was about even.
And worth saying what that win doesn't cover. Those two burgers had higher estimated environmental impact than the Big Mac. They won on taste, not on the other goals.
What about the mushroom burger, with its 94% lower impact?
It scored 0.06 against 0.93. But tasters rated it significantly lower on liking, flavor, and texture.
That figure is an estimate from aggregated databases, not a measurement of any actual kitchen's footprint. Useful for comparison, but not precise.
The nutritious burger nearly doubled the Big Mac's HEI score. Does that mean it's healthier to eat?
On paper, yes, 63.12 versus 33.71. But it's calculated from ingredients. Nobody measured health outcomes.
Right, and tasters called it bland, dry, and grainy. A healthy burger people won't finish has limited use.
And the Big Mac match, taking 7.3 million recipes?
I'd be careful there. The reference was rebuilt from four public recreations because the real formula is secret, and the match allowed quantity variation. It doesn't show the AI reproduced McDonald's recipe.
The takeaway is the trade-offs. No single burger led on taste, nutrition, and sustainability, and the model mostly reflects Western recipes.
El Pulso
- A hundred and one volunteers sat down blind to six burgers, and the question was simple and unforgiving: would people actually want to eat what the algorithm made?
- Food production strains the climate, the water supply, and public health, and the researchers wanted a faster way than slow chef-led trial and error to find alternatives people would accept.
- The model sifted a million generated recipes, scoring each for novelty, estimated appeal, environmental cost, and nutrition, then passed its best ingredient lists to a chef who turned them into real, repeatable burgers.
- Two AI-designed burgers narrowly beat the Big Mac on flavor and liking, yet the mushroom-heavy eco burger cut environmental impact by about 94% and tasters found it noticeably less enjoyable.
- No single burger managed to win on taste, nutrition, and sustainability together, and the tool was trained mostly on Western recipes, so the search for a burger that satisfies everyone remains open.
In a San Francisco restaurant, a hundred and one people quietly compared five burgers designed by a generative AI model against the most familiar burger of the modern age, the Big Mac, not knowing which was which. The effort, reported in npj Science of Food by researchers at Stanford, belongs to an old human struggle: to eat in ways that feed people well without wounding the land and water that sustain them. Two machine-made burgers edged past the classic in flavor and liking, though only by modest margins, while the designs chasing climate or health goals tended to give up some of the pleasure they were meant to carry. The result suggests that taste, nutrition, and sustainability still pull against one another, and that a recipe's future depends on whether it can persuade a palate as well as a spreadsheet.
On a single afternoon in San Francisco, 101 volunteers tasted six burgers without knowing which was which. One was a Big Mac. The other five had been designed by a generative AI model built by researchers at Stanford University. Each taster rated liking, flavor, and texture, and described what they noticed, words like meaty, smoky, earthy, and bland.
The study, published in npj Science of Food, began with a practical worry. Food production drives climate change, drains freshwater, and erodes biodiversity, while poor diets fuel malnutrition and chronic disease. Greener and healthier alternatives already exist, but people only embrace them if they taste good and feel familiar. Traditional product development depends on chefs' experience and long rounds of testing, which is slow. The team asked whether AI could shorten that path.
The model learned from 2,216 human-written burger recipes, drawn from a much larger online collection, and learned which ingredients tend to appear together and in what proportions. It then generated one million new recipes and scored them for novelty, predicted appeal, environmental impact, and nutrition. Appeal was estimated from recipe popularity, a rough stand-in for taste rather than a true measure of it. An executive chef translated the chosen ingredient lists into cooking instructions that others could follow.
The two designed for deliciousness did best. Both were judged slightly better than the Big Mac on flavor or overall liking, though the gaps were small. Their environmental footprints, however, were higher than the classic's.
The sustainability and health designs revealed the trade-offs. A mushroom-based burger cut estimated environmental impact by about 94%, but tasters found it clearly less enjoyable in liking, flavor, and texture. A bean-based version with more vegetables and whole grains scored far higher on nutrition index, yet was described as earthy, bland, dry, and grainy, and was rated lower overall.
In the end, no single burger excelled at taste, nutrition, and sustainability at once. The authors caution that the model mostly reflects Western burger traditions and that its health and environmental gains come from ingredient data rather than real health outcomes. They call for wider sensory testing and more complete models before anyone expects a machine to design the perfect burger.
At a restaurant in San Francisco, 101 volunteers each sat down to taste six burgers without knowing which was which. One of them was a Big Mac. The other five had been designed by a generative AI model built by researchers at Stanford University and the Stanford University School of Medicine. They rated each burger on overall liking, flavor, and texture using a seven-point scale, and also described what they tasted: meaty, smoky, earthy, bland.
The work, published in the journal npj Science of Food, starts from a practical problem. Food production contributes to climate change, biodiversity loss, and freshwater depletion, and poor diets lead to malnutrition and chronic disease. Sustainable, nutritious alternatives exist, but consumers accept them only if they taste good and feel familiar. Conventional product development leans on chefs' expertise and repeated testing, which is slow. The team wanted to see whether AI could speed that up.
The model worked in two stages. It was trained on 2,216 human-written burger recipes, drawn from a collection of more than half a million on Food.com and covering 146 ingredients. From those it learned which ingredients tend to appear together and in what amounts. It then generated one million burger recipes and scored them on novelty, predicted palatability, environmental impact, and nutritional quality. Palatability was estimated from recipe popularity, which the authors treat as a proxy for appeal, not a direct measure of taste. Environmental impact combined land use, freshwater eutrophication potential, scarcity-weighted water use, and greenhouse gas emissions. Nutrition was scored with the Healthy Eating Index, or HEI. An executive chef then turned each selected ingredient list into standard cooking and assembly instructions so other chefs could prepare them consistently.
In a separate test, the model was asked to find a recipe matching a reference Big Mac, without having been trained on it. Because the official formula is proprietary, the researchers built the reference from four publicly available recreations. Across ten independent runs, a match took an average of 7.3 million generated recipes. The match allowed some variation in ingredient quantities, so it does not show the model reproduced the real commercial recipe.
In the tasting, the two burgers designed for deliciousness did best. Delicious Burger 1 had slightly higher flavor ratings than the Big Mac. Delicious Burger 2 had slightly higher overall liking and flavor. The differences were statistically significant but modest, and texture scores were similar. Tasters more often called the first meaty, moist, and fatty, and the second meaty and smoky. Both, however, carried higher estimated environmental impacts than the Big Mac.
The sustainability and nutrition designs showed the cost of those goals. Sustainable Burger 1, a mushroom-based formulation, scored 0.06 on environmental impact against 0.93 for the Big Mac, about 94% lower, but tasters rated it significantly lower on liking, flavor, and texture. Sustainable Burger 2, which combined beef and mushrooms, scored 1.02, slightly worse than the Big Mac, and its sensory ratings showed no statistically significant differences from it.
The Nutritious Burger, a bean-based version with more vegetables, whole grains, and plant protein and less refined grain, sodium, and saturated fat, reached an HEI of 63.12 against 33.71 for the Big Mac. Its estimated environmental impact was roughly one-sixth of the Big Mac's. Those gains were calculated from ingredient data, not shown through health outcomes. Tasters scored it significantly lower on liking, flavor, and texture, describing it as earthy, bland, dry, soft, and grainy.
The model also produced personalized recipes based on age, sex, body composition, and activity level, including ones for an active 15-year-old male and a moderately active 70-year-old female. These were computational examples and were not tasted.
No single burger won on taste, nutrition, and sustainability together. The authors note the model mainly reflects Western burger traditions, considers ingredients and quantities but not cooking methods, and depends on aggregated environmental and nutritional databases. They say broader sensory testing and more comprehensive food-design models are needed.
Citas Notables
No single tested burger achieved the highest scores for taste, nutrition, and sustainability at once.— Study findings, Stanford researchers, npj Science of Food