Abstract
Generative artificial intelligence (AI) is becoming a groundbreaking paradigm in the field of animal genomics and is providing the possibility to take a step towards intelligent agriculture, with better data integration, predictive modeling, and biological design. This review focuses on the shift from predictive to generative modelling paradigms, examining their implications for data synthesis, biological sequence design, and integrative smart livestock systems. It provides a comprehensive overview of recent developments, applications, and challenges at the intersection of generative AI and animal genomics, as well as future directions. In doing so, it sheds light on novel opportunities and constraints specific to livestock genomics that are not adequately addressed in broader AI or human genomics studies. Thus, it bridges the gap between computational innovations and biological constraints. It initially sets the conceptual background in place by looking at the development of smart agriculture, the essentiality of animal genomics, and the development of generative model architectures in life sciences, as well as fundamental methodological aspects, including livestock genomic and multi-omics data peculiarities and the representation of biological sequences. The review then comprehensively discusses a wide range of applications such as genomic data augmentation, prediction of new genetic variants, design of protein and gene sequences, augmentation of genomic selection and trait prediction, regulatory and epigenomic modeling, accurate breeding and reproductive technologies, and cross-species genomic modeling, illustrating how generative AI is transforming genomics into something generative, enhanced through simulation. It is discussed in terms of integration into systems of smart agriculture, connections with precision livestock farming, digital twin, genomics-to-management pipelines, and sustainability-focused systems of decision-making, where the adaptive, individualized, and system-level optimization can be applied. Critical analysis of major challenges and limitations, such as heterogeneity and scarcity of data, model bias and generalization, computational and resource limitations, validation and interpretability issues, and ethical, legal, and social constraints that drove the responsible deployment are also critically reviewed. Lastly, the future opportunities are discussed, which should center on generative genome engineering, multimodal and federated modeling, species preservation, real-time interaction with smart farming technologies, and the creation of responsible and ethical AI frameworks. Overall, it is possible to state that this review makes generative AI a base technology of the new generation of animal genomics and smart agriculture, but it highlights that interdisciplinary cooperation, stringent validation, and alignment with the notions of sustainability, animal welfare, or values are necessary to realize its capabilities.