Although it is well established that land use and management affects soil microbial biodiversity, with consequences for soil-mediated ecosystem services, knowledge on human-induced variability in soil bacterial and fungal communities is more abundant for cropping than for livestock systems. In particular, grazing management affects vegetation cover and light availability to lower vegetation layers, particularly in wooded grazing ecosystems. Our objective was to evaluate the use of T-RFLP analysis to characterise soil microbial community patterns. We worked in a grazed temperate savannah in Entre Ríos (Argentina) and focused on two farms with different grazing management, resulting in different vegetation cover. At each farm, we sampled soils from four conditions forming a two shrub cover conditions (open and close) by two light availability conditions (high and low) factorial set with three replicates. T-RFLP data for bacteria and fungi were used to characterise biodiversity patterns (diversity indexes, relative abundance of operational taxonomic units and community structure). Analyses were performed, between and within farms, using nonmetric multidimensional scaling analysis, non-parametric multivariate analysis of variance and network analysis. Variation patterns of soil microbial communities were detected using T-RFLP analysis in savannah soils from these two livestock farms. Farm, light and shrub conditions were statistically significant factors shaping variation patterns. For bacterial communities, light availability and shrub cover affected biodiversity and community structure respectively, but this impact was detected only for the farm with lower grazing intensity. For fungal communities, light availability was the factor shaping community structure, also with slightly higher impact on the farm with lower grazing intensity. These results suggest that in temperate woodlands lower grazing intensity, associated with higher vegetation cover, increases bacterial