To evaluate the efficacy of passive restoration on soil seed bank and vegetation recovery, we measured the species composition and density of the soil seed bank, as well as the species composition, density, coverage, and height of the extant vegetation in sites passively restored for 0, 4, 7, and 12 years (S0, S4, S7, and S12) in a degraded grassland in desert land. Compared with S0, three more species in the soil seed bank at depths of 0–30 cm and one more plant species in the community was detected in S12. Seed density within the topsoil (0–5 cm) was five times higher in S12 than that in S0. Plant densities in S7 and S12 were triple and quadruple than that in S0. Plant coverage was increased by 1.5 times (S4), double (S7), and triple (S12) compared with S0. Sørensen’s index of similarity in species composition between the soil seed bank and the plant community were high (0.43–0.63), but it was lower in short-term restoration sites (S4 and S7) than that in no and long-term restoration sites (S0 and S12). The soil seed bank recovered more slowly than the plant community under passive restoration. Passive restoration is a useful method to recover the soil seed bank and vegetation in degraded grasslands.