Community composition data are valuable for conservation management, facilitating identification of rare native and invasive species, along with abundant ones. However, traditional capture-based morphological surveys require considerable taxonomic expertise, are time consuming and expensive, can kill rare taxa and damage habitats, and often are prone to false negatives. Alternatively, metabarcoding, employing high-throughput sequencing (HTS), can be used to assess the compositions of entire communities from environmental samples, offering a more sensitive, less damaging, and relatively time- and cost-efficient approach. However, a trade off exists between stringency of bioinformatic filtering to remove false positives and the potential for false negatives. The present study design thus employed four mitochondrial (mt) DNA assays and a bioinformatic pipeline to increase confidence in species identifications by removing false positives from several potential sources, while achieving high detection probability. Positive controls were used to calculate sequencing error, and species hits were removed that fell below the cutoff unless they occurred in multiple assays. We tested performance of our assays and this reduced-error pipeline with mock community, laboratory aquarium, and in situ field experiments to detect and identify North American freshwater fishes, on environmental (e)DNA water and bulk organism samples.