AbstractFish biodiversity can be measured by capturing and then sequencing free DNA present in water. Such environmental DNA (eDNA) methods offer an effective, noninvasive tool for species diversity measurement, although standardized protocols are not yet developed. We investigate how metrics of fish biodiversity revealed through eDNA analysis of water are influenced by sampling volume. Water samples were collected from the intertidal reef of Browse Island, a tropical, remote island in the Timor Sea. Aliquots from a single 20,700‐ml sample and multiple 2,000‐ml samples were filtered in various volumes (25–2,000 ml) across two membrane sizes (0.20 and 0.45 µm). A fish metabarcoding assay was used to characterize the fish diversity within aliquots. All samples, except one, yielded fish DNA sequences. Two hundred and nine operational taxonomic units (cf. species) representing 48 fish families were identified from the complete collection of DNA contained in all samples, comparable to the 200 fish species detected using conventional surveys at this location. Notable additions from eDNA methods were cryptic and nocturnal fish species. Nevertheless, large differences in taxonomic composition (<60% species overlap) between aliquots of identical volumes demonstrate that eDNA in seawater is patchy and that estimates of biodiversity are strongly influenced by the volume of water filtered. We suggest that eDNA studies maximize water volumes as much as logistically possible if the aim is to detect the greatest number of taxa and that species accumulation curves be provided as an indication of sampling adequacy.