Environmental DNA (eDNA) metabarcoding has been used increasingly to assess biodiversity of aquatic vertebrates. However, there still remains to be developed a sampling design of eDNA metabarcoding that can ensure high detection rates of species with minimum total survey effort, especially for large-scale surveys of aquatic organisms. We here tested whether pooling of eDNA samples can be used to evaluate biodiversity of freshwater fishes in four satellite lakes of Lake Biwa, Japan. Fish communities detected by eDNA metabarcoding of the mitochondrial 12S region were compared between the individual and pooled samples. In the individual samples, 31, 22, 33, and 31 fish lineages (proxies for species) were observed at the respective sites, within which moderate spatial autocorrelation existed. In the pooled samples, 30, 20, 29, and 27, lineages were detected, respectively, even after 15 PCR replicates. Lineages accounting for < 0.05% of the total read count of each site's individual samples were mostly undetectable in the pooled samples. Moreover, fish communities detected were similar among PCR replicates in the pooled samples. Because of the decreased detection rates, the pooling strategy is unsuitable for estimating fish species richness. However, this procedure is useful potentially for among-site comparison of representative fish communities.