Abstract: Aquatic pollutants source tracing is fundamental to effective pollution control and ecological remediation, providing critical technical support for basin-scale ecosystem management and socio-ecological governance. Conventional methods, constrained by insufficient detection sensitivity, oversimplified models, and inadequate cross-domain data integration, fail to support dynamic full-chain (source-pathway-sink) analysis of pollutant behavior. This review synthesizes recent advances in aquatic pollutants source tracing methods and emerging trends driven by artificial intelligence (AI) technologies. Bibliometric analysis and case studies identify three converging research frontiers intelligent source recognition, dynamic transport modeling, and sink-related risk early warning. AI technologies, high-dimensional data processing, nonlinear relationship mining, and dynamic simulation, are pollutant tracing frameworks across multiple dimensions. In source apportionment, deep learning architectures integrated with transfer learning frameworks demonstrate exceptional performance in multi-scale source fingerprinting under heterogeneous conditions via multimodal data fusion. For transport modeling, physics-informed neural networks and hybrid time-series prediction models enable precise prediction of cross-media (water-atmosphere-soil) pollutant trajectories. Advanced sink-risk assessment integrates multi-scale early-warning systems with mechanistic frameworks to elucidate sink formation dynamics and ecological risk propagation pathways. Future research should prioritize cross-basin generalizability, explainability of physical and chemical mechanisms, and lightweight deployment. Equally critical are collaborative platforms enabling comprehensive environmental perception and intelligent decision-making, alongside technical system standardization and ethical governance frameworks.