Nutrition experts urge scientific validation of AI-generated ingredient innovations

AI can compress discovery timelines in nutrition innovation and generate a wealth of new ideas. However, industry experts underscore that successful use of the technology requires scientific validation of AI-generated insights, high-quality data, human oversight, and clear communication on its use to customers. 

In the second installment of our series on AI in nutrition, Nutrition Insight speaks with representatives from Brightseed and Nuritas, as well as independent marketing and R&D experts, about the challenges companies face in working with the technology and how to overcome them. 

Lee Chae, Ph.D., co-founder and CEO at Brightseed, believes that AI can help the nutrition industry shift from trend-driven to evidence-driven innovation. 

“Consumers are increasingly looking for products that support proactive health, but they are also more skeptical of unsupported claims. Regulators and commercial partners are also raising expectations for substantiation and scientific defensibility.”

Dr. Nora Khaldi, founder and CEO of Nuritas, highlights what makes successful AI-led functional ingredient innovation. “Companies with clinical trial portfolios, peer-reviewed publications, and reproducible results across independent institutions will separate from those making predictive claims without validation.”

She points to data quality as a key challenge. “Predictive models are only as strong as the proprietary or experimentally-generated data that feed them. With sparse or third-party data, the model’s predictions are weaker.” 

Chae says the functional ingredient industry can raise its standard through AI by building products with stronger proof from the start.Chae says avoiding “black-box decision-making” is another challenge. “Health science teams need explainability. They need to know why a bioactive was selected, what mechanisms are implicated, what evidence supports the hypothesis, and what still needs to be tested.” 

“In this sector, AI adoption will depend on trust — and trust depends on data quality, scientific transparency, and validation.”

Evidence-driven innovation

According to Chae, the functional ingredient industry has an opportunity to raise its standard through AI by building products with stronger proof from the start, grounding discovery in high-quality data, biological plausibility, and clear evidence pathways. 

“The future of this space will not be about AI replacing scientists. It will be about giving scientists and health science teams better tools, data, and decision-making infrastructure so they can develop innovations that truly advance human health.”

“Strong science requires traceability: teams should be able to understand why an AI system surfaced a particular ingredient, what mechanisms support it, what evidence exists, and what assumptions still need to be validated.”

Chae adds that a product’s credibility depends on whether it can stand up to scientific, regulatory, partner, and market scrutiny, which means that AI-led discovery should be paired with mechanistic research, translational science, and, where appropriate, human clinical studies.

“AI should help teams ask better questions earlier, identify stronger candidates, and design more efficient validation pathways. But the standard for success should remain the same: evidence that can be proven, defended, and translated into real-world health impact.”

Transparent scientific process

Although AI can help identify patterns in predicting clinical outcomes, companies still need to consider biological complexity, argues Palak Uppal, an independent nutraceutical innovations and AI strategy expert. 

“Every human is different — their genetics, diet, and metabolism. We can’t expect the same results from an ingredient for everyone. Some people cannot digest folic acid, but they can digest folate,” she exemplifies. “That’s why relying on clinical data is one of the best outlets.” 

Khaldi urges using the same rigor on AI-discovered ingredient candidates as non-AI-identified ones before making efficacy claims.Nuritas’ Khaldi advises companies to recognize that predictive accuracy is different from clinical proof. 

“AI output is a hypothesis-generation step, not a substitute for proof. The AI-discovered ingredient candidates should pass the same rigor as non-AI-identified ones before efficacy claims are made.” 

Any ingredient business using AI should keep transparency top of mind, she adds. “A primary industry priority should be ensuring there is a clear definition of, and distinction between, ‘AI-predicted’ and ‘clinically shown or proven’ for ingredient companies, regulators, industry partners, and consumers alike.” 

Khaldi also underscores that companies should be prepared to cite exact studies, sample sizes, and outcomes that align with the claims they’re making and their promotional language. 

Data quality

The experts point to data quality as one of the key challenges in working with AI tools, as the technology is “only as strong as the data.” 

“Ingredient innovation is highly complex,” says Chae. “Bioactives do not operate in isolation; they interact with biological systems, ingredient matrices, processing conditions, dose, bioavailability, population differences, and real-world behaviors. Predicting efficacy requires more than matching an ingredient to a trend or a published claim.”

“If AI systems are trained only on generic or poorly structured data, they can produce outputs that appear sophisticated but are not scientifically actionable,” he cautions.

Uppal adds that data quality suffers because not all information is public. For example, she highlights that market data is very reliant on retailers, such as Amazon, while many ingredient companies are hesitant to share their sales data at an ingredient level because “they don’t want their sales numbers out there, as their revenue would show.”

Uppal warns that missing and incorrect data undermine AI systems across the nutraceutical sector.As a result, she says there is a lot of missing data in the nutraceutical world. 

Moreover, she adds that there is substantial incorrect data that feeds into AI systems, which is why the technology has a bad reputation. “People don’t know how to skim that data really thoroughly.” 

Human intelligence

Although AI can help speed innovation, Uppal underscores the importance of human intelligence. 

“The most successful organizations combine AI with multidisciplinary expertise — from formulation scientists, regulatory professionals, clinicians, and engineers — and then understand which data make sense and which do not.” 

She also advises companies to avoid confirmation bias. “AI should challenge assumptions, not simplify or reinforce what teams already believe.”

Jenny Mason, managing director at BDB Global, a B2B marketing agency in the nutrition and food ingredient industries, adds that AI will expose “how clear a business is about its own value.” 

She notes that leadership has to decide what deserves focus. “That requires a strong understanding of where the business has genuine authority, which customer problems it is best placed to solve, and where it has the clearest opportunity to lead.” 

“For ingredient companies, that clarity should guide the whole business: where to invest, which opportunities to pursue, how to innovate, and how to work with customers. Marketing then has an important role in making that thinking visible and meaningful.” 

Mason notes AI exposes how clearly a business understands its own value, which marketing can make visible.“The technology will continue to change quickly,” she adds. “A clear sense of direction will help businesses use each new capability with purpose and stay focused on the value they are uniquely placed to create.” 

Science-based marketing

Once a supplier has developed a new ingredient, Mason notes that customers need more than a strong data package and scientific credibility to picture what it could become. 

She highlights that successful ingredient launches make the science easy to carry through an organization. “When people can understand its value from their own perspective, the product becomes easier to discuss, support, and champion internally.” 

“Different stakeholders need different reasons to believe in the opportunity. The challenge is to find the part of the science that matters most to each audience and build the story from there.” 

Mason explains that a formulator may consider ingredient performance, stability, format, or ease of use. Meanwhile, a marketing or commercial team may want to know which consumer needs it addresses or the reason why it is relevant now. “The emphasis changes with the audience, but the evidence beneath it must not.” 

She highlights that companies need to make a distinction between product and brand-level communication. “For the individual ingredient, science has to lead. Customers need to understand what has been proven, how it performs, and where it can be used.” 

“The emotional connection is carried more by the brand — whether customers trust the business, believe in its expertise, and feel confident that it can help take the ingredient from technical potential to commercial reality,” Mason details. “A good story should make both clear without stretching what the science can support.”