Understanding the mechanistic links between environmental drivers, human disturbance, plant functional traits, and ecosystem properties is a fundamental aspect of biodiversity-ecosystem functioning research. Recent studies have focused mostly on leaf-level traits or community-level weighted traits to predict species responses to grazing and the consequent change in ecosystem functioning. However, studies of leaf-level traits or community-level weighted traits seldom identify the mechanisms linking grazing impact on leaf traits to ecosystem functioning. Here, using a multi-organization-level approach, we examined the effects of grazing on leaf traits (i.e., leaf area, leaf dry mass and specific leaf area) and ecosystem functioning across six communities of three vegetation types along a soil moisture gradient in the Xilin River Basin of Inner Mongolia grassland, China. Our results showed that the effects of grazing on leaf traits differed substantially when scaling up from leaf-level to species, functional group (i.e., life forms and water ecotype types), and community levels; and they also varied with vegetation type or site conditions. The effects of grazing on leaf traits diminished progressively along the hierarchy of organizational levels in the meadow, whereas the impacts were predominantly negative and the magnitude of the effects increased considerably at higher organizational levels in the typical steppe. Soil water and nutrient availability, functional trade-offs between leaf size and number of leaves per individual, and differentiation in avoidance and tolerance strategies among coexisting species are likely to be responsible for the observed responses of leaf traits to grazing at different levels of organization and among vegetation types. Our findings also demonstrate that, at both the functional group and community levels, standing aboveground biomass increased with leaf area and specific leaf area. Compared with the large changes in leaf traits and standing aboveground biomass, the soil properties were relatively unaffected by grazing. Our study indicates that a multi-organization-level approach provides more robust and comprehensive predictions of the effects of grazing on leaf traits and ecosystem functioning.