Under the ‘dual carbon’ goals, coal-fired power plants face severe challenges in their low-carbon transition. Key decarbonisation pathways, such as flexible retrofitting, fuel blending, and carbon capture, utilisation, and storage (CCUS), are all hindered by complex processes, conflicting objectives, and high costs. This paper aims to systematically explain how artificial intelligence (AI) can provide data-driven solutions to overcome these challenges and accelerate the low-carbon transition of coal-fired power plants. This review employs bibliometric tools to systematically screen literature and analyse research hotspots and evolving trends. The study constructs a ‘three-stage evolution model’ for AI-enabled coal-fired power plants, revealing a paradigm shift from local optimisation to autonomous evolution. The study found that the core of AI lies in transforming industrial challenges that are difficult to model precisely using traditional methods into data problems that are learnable, predictable, and optimisable. In terms of flexible operation, AI achieves combustion optimisation and stable control through high-precision prediction models and advanced controllers. In fuel blending, AI establishes intelligent combustion models to optimise boiler efficiency, emissions, and fuel supply chain management. In the CCUS field, AI permeates the entire ‘capture-transport-storage-utilisation’ chain, fundamentally reshaping its technical and economic feasibility through energy consumption optimisation, new material discovery, and efficiency improvements. AI is driving the evolution of thermal power plants from local optimisation to a plant-wide, full-life-cycle intelligent operation paradigm based on digital twins.