Global biodiversity loss is one of the most serious environmental crises of the 20th and 21st centuries. Fish diversity, an important component of biodiversity, is also in decline. The conservation of fish diversity is therefore essential for ecosystem management and the sustainable use of resources. Most current aquatic biodiversity surveys rely on conventional methods. The emergence and application of environmental DNA (eDNA) technology has triggered unprecedented innovations in aquatic organism investigation and monitoring. eDNA metabarcoding refers to the selection of appropriate universal primers for different target taxa and the identification of multiple target species in environmental samples via PCR amplification combined with high-throughput sequencing. The Zhoushan Archipelago, located in the northeastern waters of Zhejiang Province, consists of 2 085 islands with a total area of 22 216 square kilometers. Located at the confluence of the Yangtze River, the Qiantang River, the Yongjiang River, and the Taiwan Warm Current, the Zhoushan Archipelago forms a unique and complex hydrological environment, which provides a rich source of nutrients and a suitable habitat for fish. In recent years, the large-scale exploitation of marine resources and marine pollution have led to a serious loss of fish diversity near the Zhoushan Archipelago. To restore fish diversity, conservation efforts have been implemented by the government and researchers, including stricter fishing bans and ecological fish stock enhancement and release. However, the lack of reliable tools to understand the species and assess biodiversity is the most critical issue in conservation efforts. Trawl sampling and morphological identification are the primary fish diversity assessment methods. However, these conventional methods are often inefficient, time-consuming, selective, and destructive. The extensive rocky reefs in the Zhoushan Archipelago make it difficult to use conventional fishery resource survey methods efficiently. Therefore, simple and accurate methods for species diversity monitoring are needed. In this study, Xixuan Island, an offshore island in Zhoushan, was taken as a reference. Four different sampling stations were designed, and a total of 12 water samples were collected in February, May, and November 2019. The species composition and diversity characteristics of offshore fish communities around Xixuan Island were described, and the spatial and temporal differences in fish diversity were evaluated. A total of 33 fish species belonging to 12 orders, 26 families, and 32 genera were detected. Among them, Perciformes has the most abundant species, with 19 species, accounting for approximately 57.6% of all species detected. The diversity and evenness indices of the different sampling seasons were significantly different (P < 0.05), indicating that season may be one of the factors affecting the diversity of offshore fish around Xixuan Island. The results of the combined temporal and spatial analysis showed that more fish species were detected during the breeding season at sampling sites far from the main island side of Zhoushan. By comparing the results with the conventional fishery resources survey, we found a large variation in the dominant species in different seasons, which may be related to the limited sampling site numbers. Temperature and longitude may be important factors affecting the fish diversity around Xixuan Island. The result of enriching the dominant species in each season by heat trees differed significantly from those of traditional surveys, indicating that eDNA cannot completely substitute the conventional survey methods; however, it can be applied as an auxiliary method to monitor fishery resources, improve detection efficiency, and reduce interference to ecosystems. This study not only provides novel methods for the fishery resources survey in offshore Zhoushan but also provides technical support for the subsequent investigation and evaluation of fish diversity and its protection and management.