Manufacturing systems are undergoing a fundamental transition as efficiency-driven optimisation paradigms prove increasingly inadequate for meeting net-zero, resource-efficiency, and resilience objectives. Digital twins have emerged as a central enabler of this transition, offering continuously coupled physical–digital representations capable of real-time monitoring, prediction, and control. Recent advances in artificial intelligence have accelerated this evolution, transforming digital twins from static simulation artefacts into adaptive, learning-enabled systems embedded within cyber–physical manufacturing environments. However, this shift has also exposed critical challenges related to trust, interpretability, scalability, and sustainability alignment. This review provides a critical synthesis of AI-enabled digital twin research with a specific focus on manufacturing and additive manufacturing systems. It examines the progression from physics-based and data-driven twins toward hybrid AI–physics architectures that balance predictive performance with physical consistency and explainability. Beyond technical performance, the review reframes digital twins as decision-making infrastructures whose value depends on how effectively they integrate energy consumption, material efficiency, carbon intensity, and lifecycle impacts into optimisation and control logic. Particular attention is given to real-time optimisation, predictive maintenance, and intelligent asset management, highlighting persistent gaps in uncertainty propagation, cross-scale coordination, and sustainability-aware governance. The review further identifies structural barriers to large-scale industrial adoption, including data interoperability fragmentation, platform lock-in, organisational resistance, and regulatory ambiguity surrounding AI-driven decisions. Synthesising insights across domains, it argues that many current digital twin implementations remain technically sophisticated yet strategically conservative, reinforcing throughput-centred objectives rather than enabling systemic decarbonisation and circularity. The paper concludes by outlining future research directions and policy-relevant opportunities, emphasising the need for digital twins that reason across timescales, objectives, and lifecycle boundaries. By aligning manufacturing intelligence with measurable sustainability outcomes, AI-enabled digital twins can move from incremental efficiency gains toward transformative impact in net-zero and circular manufacturing systems.
Francis T. Omigbodun (2026) studied this question.