Urban form evolution and regional disparities in China over three decades of rapid urbanization

China has experienced the world’s fastest and most extensive urbanization over the past four decades, generating profound social, economic, ecological, and environmental impacts. Yet comprehensive national-scale assessments of long-term urban form evolution remain limited. This study investigates the spatiotemporal evolution of urban form across 236 Chinese cities from 1990 to 2020, using geospatial big data and nine indicators representing urban size, spatial structure, and morphological characteristics. We quantify long-term trajectories of these indicators and apply K-means clustering to identify distinct urban form evolution patterns and regional disparities. The results show that Chinese cities have experienced substantial expansion in built-up land and population, accompanied by sustained declines in population density, reduced spatial evenness of population distribution, and increasing morphological fragmentation. Meanwhile, internal structural attributes—including land use mix and road network density—improved markedly, although gains in urban greenness lagged behind rapid land conversion. Eastern cities exhibited the fastest expansion and strongest structural upgrading, whereas many central and western cities experienced slower structural improvements and more uneven population allocation. Cluster analysis revealed three dominant evolution patterns: Super Cities (9.3%), characterized by accelerated expansion and highly developed road networks and mixed land use; Sprawling Cities (41.9%), marked by rapid outward growth, low and uneven population density, lagging structural development, and increased shape complexity; and Compact Cities (48.7%), defined by limited land expansion, more stable population density, moderate structural improvement, and relatively regular urban morphology. Our findings offer a scientific basis for steering China’s ongoing urban transition toward more sustainable and resilient development.