Abstract: Air pollution poses a major threat to human health. Against the backdrop of climate change, China’s atmospheric pollution governance entered a new stage featuring the collaborative prevention and control of complex air pollution, with the pollution management paradigm undergoing a profound shift from total emission control and ambient quality control to health-oriented risk prevention. Following field observation, numerical simulation and laboratory research, the artificial intelligence (AI) has gradually evolved into the fourth pivotal research approach centered on the data-driven paradigm. To systematically clarify the research progress and application prospects in this field, this study combines bibliometric analysis with systematic review based on core databases including Web of Science and China National Knowledge Infrastructure (CNKI), to sort out relevant research achievements, evolutionary characteristics and research hotspots. The results reveal that publications concerning AI applications in atmospheric
pollution have experienced explosive growth in recent years. Research models have undergone iterative upgrades from traditional machine learning and deep learning to hybrid ensemble models. Intelligent perception, predictive forecasting, mechanism exploration, precise source apportionment and dynamic regulation empowered by AI have emerged as dominant research hotspots. Although AI can effectively address the limitations of conventional technologies, prominent bottlenecks still remain, including the lack of standardized datasets, poor physical interpretability, insufficient model generalization capacity, and low adaptability to extremely complex pollution scenarios. Accordingly, this paper proposes key future research directions for the deep integration of AI and air pollution prevention and control, aiming to provide theoretical and technical references for establishing a precise data-intelligencedriven air pollution prevention and control system.