Marker gene sequencing or “metabarcoding” is the primary sequencing approach currently used for molecular biodiversity surveys, but this approach is taxonomically limited and hampered by amplification biases. Shotgun metagenomes offer a PCR–free approach to environmental DNA (eDNA) sequencing, theoretically capturing the full taxonomic breadth of the eDNA pool. However, eukaryotic DNA is often a small component of metagenomes and it has seen limited use for metazoan biodiversity surveys. Here, we compare metabarcoding and shotgun metagenomes on a large (>200 sample size) set of marine water column eDNA samples and show that metagenomes can provide biodiversity information comparable to that of metabarcoding surveys. Moreover, as a result of biases in reference database composition, shotgun sequencing can outperform marker genes for certain taxa. Taxonomic database gaps remain an obstacle to accurate and comprehensive biodiversity surveys for both metabarcoding and shotgun metagenomes. We provide examples of taxa that may benefit from one approach over another and highlight cases of metagenomic utility.