Digital twin technology is transforming food supply chains through real-time monitoring, predictive simulation, and optimization. This study synthesizes existing research on digital twin applications in agri-food systems and identifies intellectual structures, thematic trends, and research gaps. Following PRISMA guidelines, a bibliometric analysis was conducted on 148 Scopus-indexed journal articles published between January 2019 and September 2025, using keyword co-occurrence and bibliographic coupling. Results show a sharp growth in publications since 2021, led mainly by India, Italy, and the United States, with most studies appearing in engineering and computer science journals. Keyword co-occurrence revealed four clusters: technological enablers of digital twins; supply-chain decision-making for resilience and sustainability; postharvest and cold-chain management; and human-centred digital twins for the food industry and precision agriculture. Bibliographic coupling confirmed these themes and indicated a shift from descriptive monitoring toward prescriptive digital twins that actively support decision-making. The review consolidates a fragmented field, highlights current biases such as the focus on horticulture and limited longitudinal validation, and emphasizes the need for interoperability, governance, and standardized metrics. It proposes future research directions for advancing predictive, resilient, and sustainable food systems.