Abstract: The large-scale and high-precision extraction of urban land information serves as an important basis for urban resource management and sustainable development. However, the existing extraction methods fail to effectively integrate other remote sensing data sources that can characterize urban land, compromising the reliability and accuracy of the extracted information. In response to this, targeting the Changsha-Zhuzhou-Xiangtan (CZT) City Cluster in central China, this study proposed a multi-feature coupled random forest method for urban land information extraction. The method utilizes the Sentinel-2 imagery as the primary data source, integrated with multi-source remote sensing data, including surface temperature, nighttime lights, population density, and gross domestic product (GDP). Furthermore, this study analyzed the spatial pattern of urban land in the CZT City Cluster. The results show that the total area of urban land in the CZT City Cluster was estimated to be 2 060.175 km² using the proposed method, distributed as 1 228.026 km² in Changsha, 385.174 km² in Zhuzhou, and 446.975 km² in Xiangtan. The extracted urban land was primarily distributed in municipal centers and prosperous townships within the CZT City Cluster, underscoring the radiating effect of Changsha as the provincial capital. The proposed method achieved an overall accuracy of 90.00%, with a Kappa coefficient of 0.87. Compared with methods using single-source remote sensing imagery, it represented an improvement of 3.81 percentage points in overall accuracy. Compared to existing urban land datasets, including SinoLC-1, GlobeLand30, and China Land Cover Dataset, this method effectively extracted the originally omitted and wrongly classified urban land. This study enables rapid, accurate, and large-scale extraction of urban land information, providing significant basic data support for the management, optimization, and sustainable development of land use in the CZT City Cluster.