Flash droughts, characterized by rapid onset and severe impacts, have emerged as a critical climate-induced hazard, particularly during the growing season when agricultural systems are most vulnerable. Despite growing global attention, comprehensive assessments of flash drought dynamics across climatically diverse regions remain limited. Here, we systematically investigated the spatiotemporal patterns and drivers of flash droughts in China during the growing seasons from 2000 to 2020. Using high-resolution, pentad-scale soil moisture data, we identified flash droughts based on a percentile-based threshold method, and attributed their drivers using standardized precipitation index and anomalies in potential evapotranspiration, vapor pressure deficit, and surface solar radiation downwards. To assess drought dynamics, we evaluated trends in four key flash drought metrics, namely onset speed, frequency, duration, and severity, across nine major agricultural regions using Kendall’s τ correlation coefficient. Our key findings reveal that: (1) Southern China experienced the highest flash drought frequency, particularly in the Yunnan-Guizhou Plateau (6.39 ± 0.07 events per decade), the Middle-Lower Yangtze Plain (6.26 ± 0.09), and South China (5.89 ± 0.27); (2) The Huang-Huai-Hai Plain, a vital grain-producing region, showed significant increases across all drought metrics (Kendall’s τ > 0.3, p < 0.05), contrasting with slight declines in parts of southern China; and (3) Atmospheric evaporative demand-related anomalies, namely potential evapotranspiration, vapor pressure deficit, and surface solar radiation, contributed more prominently to flash drought development than precipitation anomalies. These findings highlight the urgent need to integrate atmospheric evaporative demand factors into drought monitoring frameworks, while offering a transferable approach for flash drought assessment in other monsoon-affected regions globally.