As an irreplaceable and important ecosystem, island reef waters meet the needs of fish communities for growth, reproduction, development and protection from natural predators. They are good habitats for many species. Environmental DNA (eDNA) metabarcoding technology, which combines universal primers with high-throughput sequencing (high-throughput sequencing, HTS) technology, facilitates the amplification of eDNA from multiple target groups and has great potential in establishing or improving the assessment of species diversity. Zhoushan archipelago is the largest archipelago in China, which consisting of 2 085 islands. A total of 27 water samples were collected from nine stations in Zhoushan and its adjacent waters in May 2019 in this study to better understand the species composition of the major fish communities in Zhoushan and its adjacent waters and to monitor and protect their diversity. eDNA metabarcoding technique was used to determine the fish community composition. The results showed that 52 species of fish were detected in Zhoushan and its adjacent waters, belonging to 19 orders, 36 families and 49 genera. Four species were only annotated to genus-level. Perciformes and Scombriformes account for the highest proportion, 28.85% and 15.38%, respectively. The dominant species in different sea areas were quite different. The distribution trends of the Shannon, Simpson, and Pielou indices were basically the same. They all showed the trend as Zhoushan offshore waters > Yangtze River estuary > Zhoushan offshore waters. The variation of eDNA in different water layers could be roughly divided into three trends, and the trend of most fish species' sequence abundance between water layers was highly consistent with their habitat preference. In addition, by comparing the results with other scholars, it was found that the eDNA metabarcoding results in the same sea area at the same time varied greatly, which indicated that the repeatability of the eDNA technology was less reproducible and should be used with caution in the field of fisheries resources monitoring. Therefore, eDNA technology can only partially replace traditional survey methods at present. In the future, eDNA metabarcoding can be used as an adjunct to fishery resource monitoring to improve detection efficiency and reduce interference with ecosystems. This study can provide new ideas and methods for the study of fish communities in island reef waters.