Assessing the status and distribution of fish populations in rivers is essential for management and conservation efforts in these dynamic habitats. Currently, methods detecting environmental DNA (eDNA) are being established as an alternative and/or complementary approach to the traditional monitoring of fish species. In lotic systems, a sound understanding of hydrological conditions and their influence on the local target detection probability and DNA quantity is key for the interpretation of eDNA‐based results. However, the effect of seasonal and diurnal changes in discharge and the comparability of semi‐quantitative results between species remain hardly addressed. We conducted a cage experiment with four fish species (three salmonid and one cyprinid species) in a glacier‐fed, fish‐free river in Tyrol (Austria) during summer, fall, and winter discharge situations (i.e., 25‐fold increase from winter to summer). Each season, water samples were obtained on three consecutive days at 13 locations downstream of the cages including lateral sampling every 1–2 m across the wetted width. Fish eDNA was quantified by species‐specific endpoint PCR followed by capillary electrophoresis. Close to the cages, lateral eDNA distribution was heterogenous and mirrored cage placement within the stream. In addition to the diluting effect of increased discharge, longitudinal signal changes within the first 20 m were weakest at high discharge. For downstream locations with laterally homogenous eDNA distribution, the signals decreased significantly with increasing distance and discharge. Generally, the eDNA of the larger‐bodied salmonid species was less frequently detected, and signal strengths were lower compared to the cyprinid species. This study exemplifies the importance of hydrological conditions for the interpretation of eDNA‐based data across seasons. To control for heterogenous eDNA distribution and enable comparisons over time, sampling schemes in lotic habitats need to incorporate hydrological conditions and species traits.