Effective biomonitoring is critical for driving management outcomes that ensure long-term sustainability of the marine environment. In recent years environmental DNA (eDNA), coupled with metabarcoding methodologies, has emerged as a promising tool for generating biotic surveys of marine ecosystems, including those under anthropogenic pressure. However, more empirical data is needed on how to best implement eDNA field sampling approaches to maximise their utility for each specific application. The effect of the substrate chosen for eDNA sampling on the diversity of marine taxa detected by DNA metabarcoding has not yet been systematically analysed, despite aquatic systems being those most commonly targeted for eDNA studies. We investigated the effect of four commonly used eDNA substrates to explore taxonomic diversity: (i) surface water, (ii) marine sediment, (iii) settlement-plates and (iv) planktonic tows. With a focus on coastal ports, 332 eDNA samples from Australia (Indian and Southern oceans) and Kazakhstan (Caspian Sea) were collected and analysed by multi-assay DNA metabarcoding. Across study locations, between 30-52% of eukaryotic families detected were unique to a particular substrate and less than 6% of families were found in all four substrates. Taxonomic composition varied significantly depending on the substrate sampled implying that the suitability (and bias) of an eDNA substrate will depend on the focal taxa. These findings demonstrate that single substrate eDNA metabarcoding likely underestimates the total eukaryotic diversity. Future eDNA experimental design should consider incorporating multiple substrates or select substrate(s) best suited to the specific detection of target taxa.