Effective management of biodiversity requires regular surveillance of species, and analysis of environmental DNA by metabarcoding (eDNA) holds promise to achieve this relatively easily. Yet, taxonomic inquiries into eDNA data do need suitable molecular reference data, which is more often lacking then not. To describe the magnitude of this reference data void with the example of the fish in remote fiords of New Zealand, we explored the reliability of eDNA-derived species identifications in the context of other data sources, including Baited Remote Underwater Video (BRUV), the Ocean Biodiversity Information System (OBIS), as well as a comprehensive literature corpus. From all four data sources, we obtained a total of 116 species records (106 Actinopterygii, 10 Chondrichthyes; 59 from literature, 44 from eDNA – most of which with questionable reliability, 25 from BRUV, 25 from OBIS). Concordance of taxonomies between the data sources dissolved with lowering taxonomic levels, most decisively so for eDNA data. BRUV agreed with local biodiversity information much better and fared better in detecting regional biodiversity dissimilarities. We conclude that eDNA analysis here and everywhere will remain a powerful yet disabled tool for biodiversity management without substantial effort of generating local reference data, even after almost two decades of research on eDNA multispecies surveys.