AbstractMacroinvertebrate communities are crucial for biodiversity monitoring and assessment of ecological status in stream ecosystems. However, traditional monitoring approaches require intensive sampling and rely on invasive morphological identifications that are time‐consuming and dependent on taxonomic expertise. Importantly, sampling is often only carried out once in a year, namely during late winter–spring, where most indicator taxa have larval stages in the streams. Hence, species with divergent phenology might not be detected. Here, we use environmental DNA (eDNA) metabarcoding of filtered water samples collected in both spring and autumn from five streams in Denmark to address seasonal turnover in community composition of stream macroinvertebrates. We find that eDNA read data from the same stream sampling site clearly show different communities in spring and autumn, respectively. For three of the five streams, season even appears to be a more important factor than sampling site for explaining the variation in community composition. Finally, we compare eDNA data with a near‐decadal dataset of taxon occurrences in the same five streams based on kick sampling conducted through a national monitoring program. This comparison reveals an overlap in species composition, but also that the two approaches provide complementary rather than identical insights into community composition. Our study demonstrates that aquatic eDNA metabarcoding is useful for species detection across highly diverse taxa and for identifying seasonal patterns in community composition of freshwater macroinvertebrates. Thus, our results have important implications for both fundamental research in aquatic ecology and for applied biomonitoring.