The detection of species using environmental DNA (eDNA) relies on our capacity to identify DNA from the sampled environment. Once eDNA is released into the environment, the physical degradation of individual eDNA molecules over time directly affects our ability to detect species (eDNA decay). Therefore, interpreting eDNA data requires an explicit understanding of eDNA decay to accurately infer contemporary presence or absence of a given species in the study ecosystem. Most research to date on eDNA decay has focused on single-species assays (predominantly quantitative PCR), and thus little is known on how eDNA decay affects the interpretation of metabarcoding datasets. Here, we used eDNA metabarcoding (targeting a section of the eukaryotic cytochrome c oxidase subunit I gene) of water samples from controlled tanks to examine eDNA decay in a wide variety of marine metazoan species. After the stocked organisms were removed from these tanks, we observed a sharp decrease in amplicon sequence variant (ASV) richness within the first 48 hours. Furthermore, there was a substantial change in beta diversity between 24 and 48 hours, and after 48 hours, most stocked species became undetectable. Interestingly, the estimated decay rate for each study species, calculated using a linear regression of reads over time, differed across organisms, with up to 2–3-fold difference among species. Our results showed that, for marine temperate species, the most substantial change in eDNA metabarcoding detectability occurred within the first 48 h, and after that, eDNA from several taxa became undetectable. These findings inform the ecological interpretation of metabarcoding datasets and provide estimates of eDNA decay rate that are valuable for both simulation studies and the design of future ecological surveys.