The spatial variability of microorganisms in grasslands can provide important insights regarding the biogeographic patterns of microbial communities. However, information regarding the degree of overlap and partitions of microbial communities across different habitats in grasslands is limited. This study investigated the microbial communities in three distinct habitats from Xilingol steppe grassland, i.e. animal excrement, phyllosphere, and soil samples, by Illumina MiSeq sequencing. All microbial community structures, i.e. for bacteria, archaea, and fungi, were significantly distinguished according to habitat. A high number of unique microorganisms but few coexisting microorganisms were detected, suggesting that the structure of microbial communities was mainly regulated by species selection and niche differentiation. However, the sequences of those limited coexisting microorganisms among the three different habitats accounted for over 60% of the total sequences, indicating their ability to adapt to variable environments. In addition, the biotic interactions among microorganisms based on a co-occurrence network analysis highlighted the importance of Microvirga, Blastococcus, RB41, Nitrospira, and four norank members of bacteria in connecting the different microbiomes. Collectively, the microbial communities in the Xilingol steppe grassland presented strong habitat preferences with a certain degree of dispersal and colonization potential to new habitats along the animal excrement- phyllosphere-soil gradient. This study provides the first detailed comparison of microbial communities in different habitats in a single grassland, and offers new insights into the biogeographic patterns of the microbial assemblages in grasslands.