Air pollution, water contamination, soil degradation, solid waste accumulation, and carbon emissions are increasingly interconnected, posing common challenges to environmental engineering, including diverse monitoring targets, heterogeneous data sources, competing control objectives, and delayed management responses. This review examines the application progress of artificial intelligence in environmental fields through a three-level framework encompassing monitoring, control, and management. It systematically summarizes the roles of machine learning, deep learning, reinforcement learning, transfer learning, digital twins, and the Internet of Things in addressing multi-media environmental challenges. Current studies indicate that artificial intelligence has developed a relatively mature application basis in environmental monitoring as well as showing considerable potential for pollution-control processes. At the management level, the integration of machine learning with LCA, digital twins, and multi-objective optimization is gradually transforming environmental impact assessment from static accounting toward dynamic prediction, feedback, and decision-making. Despite these advances, key bottlenecks remain in data quality, model interpretability, field-scale validation, and the lack of cross-medium collaborative frameworks. Future research should therefore improve the reliability, generalizability, and interpretability of AI models while advancing AI from a task-specific modeling tool to a system-level decision-support technology for integrated environmental governance.

As a metal-free, visible-light-responsive two-dimensional (2D) layered semiconductor, graphitic carbon nitride (g-C₃N₄) has been extensively studied for photocatalytic water purification owing to its high chemical stability, facile synthesis, and low cost. However, single-phase g-C₃N₄ still suffers from bottlenecks including rapid photogenerated electron–hole pair recombination, low visible-light utilization efficiency, insufficient specific surface area, and a limited number of exposed surface active sites, which severely restrict its practical photocatalytic performance. Rational construction of heterojunctions with matched band alignment and intimate interfacial contact serves as an effective strategy to boost the photocatalytic activity of g-C₃N₄. This review introduces the structural characteristics and photocatalytic mechanisms of g-C₃N₄, summarizes the primary types of g-C₃N₄-based heterojunctions along with their charge transfer pathways, and highlights recent research progress in emerging contaminant degradation, pathogenic microbial disinfection, and reduction of high-valence and highly toxic heavy-metal ions for water decontamination. The review closes with critical perspectives on the rational design and practical engineering prospects of g-C₃N₄-based heterojunctions. By bridging heterojunction mechanisms with water-purification applications and engineering considerations, this review is expected to guide the rational design of high-performance g-C₃N₄-based heterojunctions and accelerate their translation toward practical water purification.