Soil water content (SWC) varies notably over time and space due to soil heterogeneity, climatic conditions, vegetation, and topography. To restore vegetation rationally in arid and semiarid regions, it is important to have an accurate and rapid estimation of annual mean soil water content (AMSWC). Research on predicting of profile AMSWC from routinely available soil parameters, however, is still underrepresented in the literature. Our objective was to explore the ability of pedotransfer functions to accurately predict profile AMSWC from routinely available soil particle-size distributions (PSDs) for a small-scale hillslope on the Chinese Loess Plateau. The SWCs in soil profiles from 0- to 400-cm depths were measured regularly at 30 monitoring sites on a typical hillslope during the growing seasons in 2014 and 2015. A total of 750 soil samples were taken from the same soil layers (0-400 cm) of these sites and their PSDs were determined using a laser diffraction technique. The results indicated that AMSWC increased exponentially (P < 0.001) with the clay (R-2 = 0.632), silt (R-2 = 0.480), and fractal dimensions (R-2 = 0.616), however, they decreased exponentially (P < 0.001) with the sand content (R-2 = 0.572). Using clay content as a predictor variable, the model established separately for each 100 cm depth soil sample resulted in a higher R-2 (0.915), lower mean error (ME, 0.120%), and lower root mean squared error (RMSE, 0.999%), than the model of the whole 0- to 400-cm soil depth (R-2 = 0.830, ME = 0.199%, RMSE = 1.262%). These results showed that profile AMSWC at hillslope was able to be well-estimated from routinely available soil clay content data. The model established separately for each 100-cm depth soil sample performed well at predicting AMSWC on a small-scale hillslope on the Chinese Loess Plateau. Our study proposed a precise and rapid method for estimating profile AMSWC at hillslope, serving for the optimization of soil water management and vegetation construction in dry regions.