Palatability has long been one of the most consequential — and resource-intensive — variables in pet food development. Animal-based testing remains the industry standard, but Ana Rita Monforte, Ph.D., global flavor and data sciences manager at AFB International, is working to change that equation. At Petfood Forum 2026, she will present a framework that uses AI to predict palatability performance before a single animal trial takes place.
"The key takeaway is that we can start estimating palatability performance before animal trials by using structured historical data and predictive models," Monforte said. "Palatability is not driven by a single ingredient but by complex interactions between formulation, processing and the food matrix."
Her session, "Leveraging AI to predict pet food palatability: Integrating recipe, analytical and pet data," is scheduled for Tuesday, April 28.
From siloed data to decision-quality signals
The approach centers on converting historically fragmented palatability data into a structured, usable framework. By linking formulation variables, processing conditions and performance outcomes, the AI system generates predictive signals that can inform decisions earlier in the development process — before in vivo exposure.
"For the pet food industry, this approach provides a more systematic way to design palatable products," Monforte said. "Instead of relying only on empirical testing, predictive models can help identify promising formulations earlier, prioritize prototypes, and make better use of animal trials."
Beyond accelerating development timelines, the framework constrains the design-of-experiments search space, supports ingredient selection during formulation, and enables the design of next-generation palatants tuned for specific recipe architectures and processing conditions.
For AFB International, the value is both strategic and operational. "By harmonizing data and linking formulation, processing, and performance, we create a system that supports better decision-making," Monforte said. "The goal is not to replace scientific expertise, but to strengthen it with better data and analytical tools."
A shift in how development gets done
Monforte frames the broader significance of this work as a shift in development philosophy, from intuition-led iteration to evidence-assisted design. As formulations grow more complex, she argues, data-driven tools become less optional and more essential.
"As pet food formulations become more complex, data-driven tools can help guide development, reduce trial-and-error, and focus testing on the most promising solutions," she said.
Animal trials are not eliminated in this model — they are repositioned, Monforte noted. Rather than serving as broad exploratory tools, they become confirmation steps for well-designed prototypes, concentrating resources where they matter most.
What comes next
Looking ahead, Monforte anticipates predictive approaches becoming more deeply integrated with formulation and processing design. Future models will increasingly account for factors such as kibble structure, extrusion conditions and ingredient variability.
"Animal trials will remain essential, but they will focus more on confirming well-designed prototypes rather than broad exploration," she said. "This shift toward more targeted, data-informed development will become an important competitive advantage in the pet food industry."
Petfood Forum and Petfood Essentials show dates are April 27-29, 2026, in Kansas City, Missouri, U.S. To register or stay informed on the latest event developments, go to PetfoodForumEvents.com.
