Freshwater biomonitoring programmes routinely sample aquatic macroinvertebrates. These samples are time-consuming to collect, as well as challenging and costly to identify reliably genus or species. Environmental DNA (eDNA) metabarcoding has emerged as a surrogate to traditional collection techniques and has been used in whole-community approaches across several taxa and ecosystems. However, the usefulness of eDNA-based detection of freshwater macroinvertebrates has not been extensively explored. Few studies have directly compared bulk sample and eDNA metabarcoding at a local scale to assess how effective each method is at characterizing aquatic macroinvertebrate communities. Here, we collected both eDNA and kicknet samples at the same sample transect locations across nine different streams in southern Ontario, Canada. We observed minimal overlap in community composition between these paired samples. Bulk tissue metabarcoding resulted in a greater proportion of sequences belonging to metazoan taxa (over 99%) than eDNA (12%) and had higher OTU richness for macroinvertebrate taxa. We suggest that degenerate primers are not effective for eDNA metabarcoding due to the high degree of nontarget amplification and subsequently low yield of target DNA. While both bulk sample and eDNA metabarcoding had the power to detect differences between stream communities, eDNA did not represent local communities. Bulk tissue metabarcoding thus provides a more accurate representation of local stream macroinvertebrate communities and is the preferred method if smaller-scale spatial resolution is an important factor in data analyses.