Deception Island is a volcanic island located southwest of the South Shetland Islands in Antarctica with more than 20 eruptions over the last two centuries. Under the significantly influence of the island's active volcanoes, the adjacent water temperature is high, and the concentrations of iron, manganese, silicon, and other elements are significantly greater, with rich biological resources. Environmental DNA (eDNA) analysis constitutes a promising tool to study the distribution and diversity of fishes. This technique is based on organisms leaving DNA in the environment, which can be extracted and sequenced to identify the species from which it originates. This study explored the potential use of eDNA metabarcoding from seawater samples to detect fish species diversity in Deception Island. The fish species composition and diversity characteristics in the waters of Deception Island were detected by high-throughput sequencing of eDNA in this study. Overall, a total of 31 species (1 order 6 families 23 genera) of Antarctic fishes were detected in the environmental samples from five sampling stations around Deception Island, and most of them were identified in previous surveys of Antarctic fishes using traditional methods. Some rare fish which were considered difficult to be caught by traditional fishing methods were detected by eDNA, such as Chionobathyscus dewitti, Akarotaxis nudiceps. OTU abundance was ranked from the largest to the smallest in taxa at the family level, with Channichthyidae and Nototheniidae accounting for the highest proportions. The dominant fish species were Champsocephalus gunnari, Notothenia rossii, Pseudochaenichthys georgianus, Bovichtus sp1., Bovichtus sp2., Pagothenia borchgrevinki, C. dewitti, Notothenia coriiceps, Eleginops maclovinus. Among the identified fish species from eDNA analysis, C. gunnari and N. rossii had the highest abundance (53.52% and 28.27%). The sampling stations were divided into nearshore and distant stations according to the distance from Deception Island or South Shetland to evaluate the diversity of fish communities in different habitats. The alpha diversity index of fish communities was calculated using Chao1, ACE index to characterize richness, Shannon and Simpson diversity index, and Pielou’s evenness. Moreover, the overall composition pattern of the data was analyzed by PCoA. The α and β diversity significantly differed between nearshore and distant sites, but the fish species composition was similar between distant sites. This research demonstrated the utility of eDNA for species diversity in fish communities, knowledge of which was essential for standardized monitoring of Southern Ocean's biodiversity. The eDNA technology was less intrusive to the environment and could provide rapid detection of fish species in the water of Deception Island, complementing to traditional survey methods. In addition, our findings can provide data support for fish diversity monitoring and fisheries resource management and conservation in Deception Island.