Grassland degradation is a retrogressive succession of grassland vegetation, which leads to the loss of biodiversity and the degradation of ecosystem functions. Fencing is an effective management practice to restore grassland ecosystem function. Microorganisms, as the critical component of soil, play important roles in maintaining the stability and function of grassland ecosystems. However, it is unclear how fencing affects prokaryotic communities in different grassland soils. In this study, two treatments of fencing and grazing were set up to investigate the effects of fencing on soil prokaryotic diversity and community structure for three grassland types (meadow, steppe, and desert grassland) in degraded grasslands of the Tibetan Plateau, using Illumina HiSeq high-throughput sequencing technology. The results demonstrated that fencing did not significantly influence the prokaryotic richness, Shannon diversity and evenness in meadow soils, but significantly increased in steppe soils (P<0.05), and slightly decreased in desert grasslands (P=0.086, 0.072, and 0.099). In fencing grasslands, prokaryotic richness, Shannon diversity and evenness negatively correlated with mean annual temperature, aridity index, and pH (P<0.01), but positively correlated with mean annual precipitation, dissolved organic carbon, aboveground plant biomass, and plant diversity (P<0.01). In contrast, soil prokaryotic richness, Shannon diversity and evenness negatively correlated with mean annual temperature and aridity index (P<0.05), but prokaryotic microbial richness and Shannon diversity did not correlate with all soil physicochemical and vegetation properties in grazing grasslands. Redundancy analysis (RDA) showed that significant differences were detected in the soil prokaryotic community structure across the three grassland types, and gradually shifted along the transition of meadow, steppe, and desert grassland (P<0.001). Variance partitioning analysis (VPA) further revealed that the prokaryotic community structure variations were mainly driven by mean annual temperature, mean annual precipitation, aridity index, and pH. Fencing significantly altered the soil prokaryotic community structure at some sample sites in different grassland types. The dominant phyla were Actinobacteria, Proteobacteria, and Acidobacteria in the three grassland types. Actinobacteria showed the highest relative abundance in desert grassland soils, while Proteobacteria and Acidobacteria showed the highest in meadow soils. In addition, there were no significant differences in the relative abundance of soil prokaryotic taxa between different types of fencing and grazing grasslands. This study showed that the prokaryotic community in different grassland soils responded differently to fencing. This provides data support for the development of local condition-specific management strategies for grasslands and offers theoretical support for the restoration of degraded grasslands.