Anthropogenically forced changes in global freshwater biodiversity demand more efficient monitoring approaches. Consequently, environmental DNA (eDNA) analysis is enabling ecosystem-scale biodiversity assessment, yet the appropriate spatio-temporal resolution of robust biodiversity assessment remains ambiguous. Here, using intensive, spatio-temporal eDNA sampling across space (five rivers in Europe and North America, with an upper range of 20–35 km between samples), time (19 timepoints between 2017 and 2018) and environmental conditions (river flow, pH, conductivity, temperature and rainfall), we characterise the resolution at which information on diversity across the animal kingdom can be gathered from rivers using eDNA. In space, beta diversity was mainly dictated by turnover, on a scale of tens of kilometres, highlighting that diversity measures are not confounded by eDNA from upstream. Fish communities showed nested assemblages along some rivers, coinciding with habitat use. Across time, seasonal life history events, including salmon and eel migration, were detected. Finally, effects of environmental conditions were taxon-specific, reflecting habitat filtering of communities rather than effects on DNA molecules. We conclude that riverine eDNA metabarcoding can measure biodiversity at spatio-temporal scales relevant to species and community ecology, demonstrating its utility in delivering insights into river community ecology during a time of environmental change.