Artificial intelligence in carbon capture, utilisation, and storage: A systematic review of methods, data, evaluation practices, and deployment readiness

Carbon Capture, Utilisation, and Storage (CCUS) is an important strategy for reducing industrial carbon dioxide (CO 2 ) emissions and supporting global decarbonisation. However, large-scale CCUS deployment is hindered by complex subsurface processes, heterogeneous data, monitoring uncertainty, operational inefficiencies, and limited decision-support capabilities. This systematic literature review (SLR) examines applications of artificial intelligence (AI) across the CCUS value chain. Following a PRISMA-aligned search, the review synthesises 44 eligible publications published between 2016 and 2026. It examines publication trends, application domains, AI methods, data sources and sensing modalities, evaluation practices, reported limitations, and future research directions. Conventional machine learning is the dominant methodological category, accounting for 45.5% of the corpus, followed by deep learning, hybrid AI, physics-informed AI, and generative AI. Applications include capture-process and materials optimisation, reservoir characterisation, CO 2 plume and pressure forecasting, leakage detection, monitoring, and operational planning. However, much of the evidence is derived from simulations, laboratory experiments, secondary sources, or individual case studies. Key barriers to industrial adoption include scarce benchmark and field datasets, weak cross-site generalisation, limited explainability and uncertainty quantification, high computational demands, and poor integration with operational workflows. The resulting research roadmap prioritises standardised benchmarks, field-scale validation, physics- and uncertainty-aware learning, multimodal digital twins, foundation models, and generative AI. These priorities can support trustworthy, scalable, and field-deployable AI systems for CCUS.