DNA metabarcoding is a promising method to increase cost and time efficiency of marine monitoring, providing that the impact of methodological choices on the reliability and reproducibility of results are well understood. Here, we investigated the impact of primer choice, DNA source (bulk DNA or eDNA from the ethanol preservative) and morphological traits (body size and body skeleton) on species detection in four distinct macrobenthos communities from the North Sea. We generated a reference database with COI sequences for macrobenthos from the North sea and applied DNA metabarcoding using five COI primer sets. At most 22% of the ASVs were assigned taxonomy at the phylum level, despite the availability of a nearly complete reference database. However, the unassigned ASVs represented only a small fraction of the total reads (13%). The Leray primer set outperformed the four other primer sets in the number of non-chimeric reads and species detected, and in the recovery of beta diversity patterns. Community composition differed significantly between bulk DNA and eDNA samples, but both sample types were able to differentiate the four communities. Importantly, the probability of detecting a species in the eDNA from the ethanol preservative was significantly lower than for bulk DNA for macrobenthos species with small to medium body size and for species with chitine or CaCO3 in their skeleton. Detection in the bulk DNA samples was not affected by the traits investigated, indicating that monitoring of macrobenthos species will be most robust when using bulk DNA as template for metabarcoding.