Abstract Shallow velocity structure surveys are very important for urban seismic hazard monitoring and risk assessment. Ambient-noise tomography provides an ideal way to obtain urban fine structure. In this study, we obtained a high-resolution 3D VS model of the metropolitan areas of the Pearl River Delta (PRD) using the ambient-noise double-beamforming method with a dense nodal array. The new model reveals shallow structures that correlate well with surface geological features, with low-velocity anomalies in fault depressions and high-velocity anomalies in fault uplifts. Our findings reveal detailed fault geometries and basin characteristics of the PRD. The Guangzhou–Conghua fault emerges as a prominent velocity boundary, playing a significant role in controlling the development and subsidence of the Longgui basin. The Xinhui–Shiqiao fault and Shougouling fault are identified as major faults that control the formation and evolution of depressions in the PRD. The basin structures in the PRD are classified as semigraben basins controlled by synsedimentary faults. The long axes of the sub-basins align with the strike of the major faults, and the deposit centers are located in close proximity to these faults. Furthermore, our investigation reveals low-velocity anomalies along the faults, suggesting the existence of pre-existing faults facilitating heat transfer and fluid/melt migration from the deep crust. Our results provide new constraints on the geometric structure of the sedimentary basins and fault systems in the PRD area, thereby contributing to urban seismic hazard assessment and offering valuable insights into potential geothermal resources.