The Chinese Loess Plateau (CLP) is prone to adverse effects from drought, especially the widespread creation of a dried soil layer (DSL), a problem intensified by revegetation under the Grain for Green Program. Using 13-year soil moisture (SM) data, we compared soil water consumption by planted shrubs (Korshinsk peashrub, KOP), planted grass (purple alfalfa, ALF), and natural grass (NAG) from 2004 to 2016 in the CLP. To assess the soil water recovery processes, long-term (30 years) SM dynamics were simulated using the simultaneous heat and water (SHAW) model based on field data and local meteorological data under two scenarios (A: converting KOP to NAG and B: converting ALF to NAG). The results showed that the decline rates of SM in 1-4 m profiles for NAG (24.0-29.8%) were much lower than those for KOP (47.6-51.4%) and ALF (48.8-50.2%) during the 13-year growth period. Modelling SM dynamics at depths of 1-4 m for 30 years showed that SM gradually increased and that the DSL prevalence could be reduced under scenarios A and B. The complete elimination of DSL requires at least 6 years at 1-4 m under scenario A, 13 years at 2-4 m, and 22 years at 1-2 m under scenario B. Soil water restored to local stable soil water levels requires approximately 19, 13, and 15 years in the 1-2 m, 2-3 m, and 3-4 m profiles, respectively, under scenario A. Soil water recovery will take approximately 28 years in the 2-3 m profile and 27 years in the 3-4 m profile under scenario B. Our results enhance the understanding of the soil water depletion and recovery processes under different vegetation types and can could be used to provide sci-entific guidance for sustainable ecological restoration in the CLP.