The global food system faces high pressure to sustain a growing population amid climate constraints and shifting consumer demands, making the traditional trial-and-error development methodologies inadequate. Artificial Intelligence (AI) has transitioned from a simple optimization tool into a structural enabler across the entire food chain. This review examines the integration and evolution of computational architectures in food technology between 2006 and 2026, tracing the paradigm shift from the early fuzzy logic and rule-based systems to modern deep learning and generative frameworks. This review highlights breakthroughs achieved over the last five years, demonstrating how Graph Neural Networks, Transformers, and Variational Autoencoders and other architectures are accelerating the in silico discovery of bioactive ingredients, predicting complex molecular flavors, and autonomously synthesizing optimal culinary formulations. The transition to Industry 5.0 is also explored, emphasizing the integration of collaborative robotics, process-level digital twins, and federated learning to enable autonomous manufacturing and privacy-preserving precision nutrition. Finally, this review addresses critical barriers to commercialization, including severe data fragmentation, the “Innovation Paradox” in fundamental academic research, and the urgent need for multidisciplinary teams capable of translating digital predictions into physically stable, strictly regulated food matrices.
Fernandes et al. (Sat,) studied this question.