In the world of big food, artificial intelligence is nothing new.
McCormick, which owns brands including Frank’sRedHot, Cholula and Old Bay, has been using AI in flavor development fornearly adecade, with the company saying its development timelines have been cut by 20% to 25%, on average, byidentifyingpromising flavor combinations and narrowing down which ideas are worth testing in physical prototypes.
It’s a similar story at Unilever, where AI is deeply embedded across food research & development, with systems able to test thousands of recipes digitally in seconds and get to viable concepts with fewer physical trials. Unilever’s Knorr Fast &FlavourfulPaste, as an example, was developed inroughly halfthe usual time. On the packaging side of the business,AI modeled how formulations behave in Hellmann’s Easy-Out squeeze bottle — which the company says saved months of physical lab work.
All the way back in 2017, a team from Google Brain (which is now part of DeepMind) used AI to help create a recipe for the “perfect” chocolate chip cookie.
But even as AI is increasingly shaping how food companies decide what ends up on grocery store shelves, the food companies are quick to stress that AI is not taking over the kitchen.
“Human creativity and judgment lead the way, and AI is a tool to help us amplify our impact,” said AnnemarieElberse, head of ecosystems, digital and data for foods R&D at Unilever.
“These tools help inspire our flavor scientists’ creativity,” Anju Rao, McCormick’s chief science officer, told CNBC.Rao emphasized that AI functions as a co-creation tool, not a replacement for humanexpertise. “Our greatest asset will always be our people who bring global perspectives, flavor expertise and human creativity to the table,” she said.
As a growing ecosystem of startups position AI as a way to approximate and predict sensory outcomesusing large datasets to model how consumers might respond to new food products before they are physically tested, it’s not clear how successful their efforts will be in cracking the code in the test kitchen.Companies including Zucca, Journey Foods,NielsenIQ,and AKA Foods market their platforms as “virtual sensory” or AI-powered systems designed to digitally screen recipes, suggest formulation changes, and predict consumer liking before physical prototypes are made.
These companies are promising much of what the food giants say they’ve been doing already: creating systems that can reduce the size of traditional taste panels, lower the risk of failed launches and compress product development cycles byidentifyingpromising concepts earlier in the process.Industry analysts estimate the global market for artificial intelligence in food and beverages will grow from roughly$10 billionin 2025 to more than$50 billionby 2030, driven by rising investment in data-driven product development, automation, and personalization.
But some early food AI pioneers have already moved on. McCormick’s early AI work was developed in partnership with IBM, which previously explored AI-driven food projects such as Chef Watson. An IBM spokesman said in a statement the company is “not actively focused in this area anymore.”
Behind the marketing language, food scientists who have tested these platforms say the technology is still early — and that many of the claims are as much about attracting capital as replacing humanexpertise.
Brian Chau, a food scientist and founder offood science and foodsystemsconsultancyChau Time, said many AI food startups are still in the data-collection phase, working to aggregate enough real-world information to make their models meaningfully predictive.
“I think all the AI companies coming out are, to some extent, overstating what they can do — that’s true of most startups,” Chau said. “They need to attract investors, they need to build datasets, and they need real industry partners before any of this really works at scale.”
Chau said most current platforms resemble large language models trained on existing recipes, manufacturing data,and consumer trendsrather than systems capable of independently generatingviablenew products.”When I tested one platform, the output was basically what you’d get from any general AI system,” he said. “There wasn’t much added value without proprietary data from real companies.”
In his view, the technology’s long-term potential depends on whether startups can secure partnerships with large food manufacturers willing to share internal formulation data — something many companies are reluctant to do because of intellectual property concerns.”Without big industry players feeding real data into these systems, it’s very hard for them to become truly predictive,” Chau said. “It’s a numbers game.”
Where AI food science still falls short
From a scientific standpoint, researchers say the biggest obstacle is not computing power —it’sbiology.
Dr. Julien Delarue, a professor of sensory and consumer science at the University of California, Davis, said expectations around AI-driven sensory tools may be inflated by misunderstandings about what AI can realistically model.”I would say there is probably a little bit of hype,” Delarue said. “It doesn’t mean that AI is not useful,it’s just not what people expect from it.”
While AI can help analyze chemical data and improve efficiency in food development, Delarue said trying to predict how people will perceive complex flavorsremainsfundamentally limited.”Trying to predict what people will perceive from a complex mixture of compounds — the answer is no,” he said.
One of the core challenges, he explained, is that human sensory perceptionis inherentlyvariable. People perceive the same chemical compounds very differently depending on genetics, culture, experience,and even personal history.”There is no such thing as the average consumer,” Delarue said. “Trying to predict what the ‘average’ person may perceive is probably a dead end.”
To unlock this limitation, Delarue says, we would need much more data than we currently have — access to data at the individual level, knowing what each person or groupactually perceives. “Andthat’sa huge task,” he added.
That variability makes it difficult for any model — human or machine — to serve as a universal proxy for taste, he said.
Even the companies building these tools emphasize that human judgmentremainscentral.
David Sack, founder of AKA Foods, said his company’s platform is designed to organize internal R&D knowledge,not replace food scientists or sensory experts.”Food R&D teams sit on large amounts of valuable knowledge, from past formulations and sensory data to tacit know-how held by individuals,” Sack said. “But it’s often fragmented and difficult to reuse systematically.”
Why humans will remain the tastemakers
AKA’s platform helps teams test ideas digitally beforecommitting tophysical trials, allowing scientists to focus on the most promising formulation paths.”It does not replace food scientists or sensory experts,” he said. “Ultimately, humansdefine the goals, constraints,and success criteria. Sensory experts design and interpret panels. Scientists decide what to test and what to launch. AI can reduce the number of tests needed, but it does not eliminate the need for real human tasting or validation. Humans will always need to be in the loop when the end consumer is human,” he said.
“Consumers decide with their palate whether they like a product,” said Jason Cohen, founder and CEO of Simulacra Data, a company that uses AI to analyze sensory and consumer data. “We still start with real human sensory data. AI just helps us extrapolate insights faster and cheaper.”
Cohen, who also foundedAnalytical Flavor Systems, which wasacquiredin 2025 byNielsenIQ, said AI is most useful foridentifyingoff-flavors, narrowing formulation options, and prioritizing which ideas are worth testing,not for replacing humanperception.
Chausays large food companies are uniquely positioned tobenefitfrom AI-driven tools because they already control vast amounts of proprietary formulation, sensory, and manufacturing data — something most small brands are still trying to build.
Delarue thinks the real value of AI within the food industry will be in efficiency not creativity — helping researchers analyze data faster, manage complexity, andoperateunder increasing constraints around health, sustainability,and cost.”Designing food today is much more challenging than before,” he said. “Youdon’tjust want to make food that people enjoy. You need to make food that is healthy, sustainable,and affordable. AI gives us more power to handle that complexity.”
But when it comes to taste itself, humans are still the reference point.”Consumers will always be the ones who decide what tastes good,” he said. “Not machines.”
About The Author
