Introduction Accurate, long-term mapping of human settlement footprints is fundamental to understanding urbanization, human-environment interactions, and sustainable development. Existing medium-resolution products primarily capture the built-up core of settlements and often overlook the heterogeneity of land cover types within these settlements. In this study, we introduce the China Settlement Footprint 1990-2020 (CSF30) dataset, the first nationwide, long-term product explicitly designed to depict complete settlement entities from Landsat imagery. CSF30 is generated by implementing our previously proposed Entity-Based Image Analysis (EBIA) framework with a deep learning (DL) architecture. Applied to Landsat imagery, this approach generated settlement maps for China at five-year intervals from 1990 to 2020. Comprehensive validation using multiple independent datasets shows that CSF30 achieves a mean F1-score of 0.91 throughout the study period. Compared with other settlement products, CSF30 demonstrates clear advantages over existing medium-resolution products in capturing the spatial footprint and temporal evolution of settlements, particularly in rural and peri-urban areas. Furthermore, at the national scale, CSF30 revealed that settlements within 2020 urban areas expanded at a rate of 0.27 million hectares (mha) per year. For settlements outside 2020 urban areas, they exhibited diverse evolutionary pathways, including expansion, disappearance, and emergence with a rate of 0.25, 0.01, 0.1 mha per year, respectively. CSF30 provides a critical, high-quality baseline for research on China’s urbanization trajectory, ecological and environmental impact assessment, and progress toward Sustainable Development Goals (SDGs). Data description This product is generated using Landsat collection 2 surface reflectance images. The data contains seven GeoTIFF files representing the China's settlement footprints in seven mapping years from 1990 to 2020 at five-year intervals. A pixel value of 1 represents a settlement, while a pixel value of 0 represents a non-settlement. Acknowledgments This study was supported by the National Natural Science Foundation of China (no. 42401474), the Otto Poon Research Institute for Climate-Resilient Infrastructure (no. N-ZH8Q) and the Hong Kong Polytechnic University (nos. 4-ZZVP and 1-WZC2).