Abstract: Benthic community structure and diversity are sensitive indicators of riverine ecological quality. With the rapid uptake of environmental DNA (eDNA) in biodiversity monitoring, eDNA-based assessments of benthic diversity have become an important tool for disentangling aquatic processes and environmental drivers. Using eDNA metabarcoding of sediments from 20 sites along the Guizhou reach of the Wujiang River, we assessed benthic community diversity, community-structure differences and their relationships with environmental factors, and to evaluate community stability via co-occurrence networks. Results identified 107 species in 82 genera, 51 families, 29 orders, 12 classes, and 4 phyla. Upstream sites showed higher diversity and evenness; downstream Chao1 richness was lower, with no significant among groups differences. β-diversity showed a clear spatial gradient dominated by few taxa: the top five explained 66.42% of dissimilarity. Mantel tests identified COD_Mn as the main gradient explaining community turnover; COD and V were marginal. Networks were sparser yet structurally robust upstream, and tighter, hub-dependent, and fragile downstream.