Elucidating how different vegetation types respond to changes in soil moisture (SM) profiles is crucial for optimizing water management and guiding ecological restoration. Here, ECMWF ReAnalysis 5 (ERA5)-Land SM and satellite-derived normalized difference vegetation index (NDVI) from 2001 to 2020 were combined to quantify spatiotemporal trends, vertical SM-NDVI coupling, stability, and coordinated change patterns across vegetation types and hydroclimatic zones in China. SM decreased in over 40.85% to 53.49% of the land area, with the strongest declines in deep layers (28 to 289 cm) in arid and semiarid transition zones and intensively cultivated regions. In contrast, NDVI increased in over 66.31% of vegetated areas, reflecting large-scale greening driven by ecological restoration and agricultural intensification, while grassland, cropland, and desert ecosystems remained highly drought sensitive. Most vegetation types showed a statistically significant positive response of NDVI to shallow SM (0 to 28 cm), whereas the influence of deep SM was more complex: grassland, cropland, and desert exhibited predominantly negative correlations, while woodland and shrubland maintained relatively stable relationships across depths. Two dominant coordination patterns were identified: "greening-drying" (34.47% to 45.84%), concentrated in arid and agricultural areas, and "greening-wetting" (38.17% to 44.61%), prevalent in humid and transitional zones. Woodland, shrubland, and marshes were mainly greening-wetting, whereas grassland, alpine, and cropland systems were mostly greening-drying. These findings indicate that the recent greening in water-limited regions partly relies on unsustainable deep SM depletion, underscoring the need to explicitly incorporate deep SM constraints into ecological restoration and water resource management.