Aquatic environmental DNA (eDNA) surveys are transforming how we monitor marine ecosystems. The time-consuming pre-processing step of active filtration, however, remains a bottleneck. Hence, new approaches that eliminate the need for active filtration are required. One exciting prospect is to use filter-feeding invertebrates to collect eDNA. While proof-of-concept has been achieved, side-by-side comparative studies to investigate the similarity between eDNA signals obtained using standard filtration approaches and those obtained from filter feeders are essential. Here, we investigated the differences among four eDNA sources (water bivalve gill-tissue sponges and ethanol in which filter-feeding organisms were stored) along a vertically stratified transect in Doubtful Sound, New Zealand using three metabarcoding primers targeting fish and vertebrate diversity. We used concurrent SCUBA diver observations to validate eDNA detection results. Combined, eDNA sources detected 59 vertebrates, while divers observed eight fish species. There were no significant differences in alpha and beta diversity between water and sponge eDNA and both sources were highly correlated. Vertebrate eDNA was successfully extracted from the ethanol in which sponges were stored, although only a reduced number of species were detected. Bivalve gill-tissue dissections, on the other hand, failed to reliably detect eDNA. While additional research into filter feeder eDNA accumulation efficiency is essential, our results provide strong evidence for the potential of using marine sponges becoming an efficient tool for eDNA based marine biodiversity surveys.