An improved understanding of increased human influence on ecosystems is needed for predicting ecosystem processes and sustainable ecosystem management. We studied spatial variation of human influence on grassland ecosystems at two scales across the Qinghai-Tibetan Plateau (QTP), where increased human activities may have led to ecosystem degradation. At the 10 km scale, we mapped human-influenced spatial patterns based on a hypothesis that spatial patterns of biomass that could not be attributed to environmental variables were likely correlated to human activities. In part this hypothesis could be supported via a positive correlation between biomass unexplained by environmental variables and livestock density. At the 500 m scale, using distance to settlements within a radius of 8 km as a proxy of human-influence intensity, we found both negatively human-influenced areas where biomass decreased closer to settlements (regions with higher livestock density) and positively human-influenced areas where biomass increased closer to settlements (regions with lower livestock density). These results suggest complex relationships between livestock grazing and biomass, varying between spatial scales and regions. Grazing may boost biomass production across the whole QTP at the 10 km scale. However, overgrazing may reduce it near settlements at the 500 m scale. Our approach of mapping and understanding human influence on ecosystems at different scales could guide pasture management to protect grassland in vulnerable regions on the QTP and beyond.