Understanding the spatial distribution patterns of aquatic environmental DNA (eDNA) is essential for standardizing eDNA sampling protocols and for ensuring the reliability and comparability of aquatic biodiversity monitoring and ecological assessments across studies. Deep reservoirs often exhibit pronounced vertical thermal stratification, and differences in the distribution of aquatic organism eDNA among water layers may affect its applicability in aquatic ecological monitoring. However, knowledge of the vertical distribution of benthic macroinvertebrate eDNA in deep reservoirs remains limited, and no consistent conclusions have yet been reached. In this study, four deep reservoirs in Wenzhou, China, were investigated in October 2023. A total of 78 independent water eDNA samples were collected from the surface, middle, and bottom layers at 26 sampling sites, along with four field negative controls; 10 physicochemical variables of surface water were measured simultaneously. After sequencing, we obtained the raw sequences. Following bioinformatics analysis and quality control (removal of low-quality sequences and contaminant/chimeric sequences), and species annotation based on the Chinese Freshwater Macroinvertebrate Barcode Library (CMBL), we obtained 31,486 valid benthic macroinvertebrate sequences, which were clustered using a 97% sequence similarity threshold to yield 255 operational taxonomic units (OTUs), representing 3 phyla, 6 classes, and 19 orders, with Arthropoda being the dominant group (221 OTUs). Sequence abundance and OTU richness were slightly higher in the surface layer than in the middle and bottom layers, although the differences among water layers were not statistically significant. No significant differences in overall benthic macroinvertebrate community composition were detected among water layers, whereas the composition of layer-specific OTUs differed significantly. The Shannon and Simpson diversity indices were higher in the surface layer than in the bottom layer, with a significant difference detected for the Simpson index. Water-quality assessments based on BMWP scores derived from eDNA communities in different water layers were generally consistent with the surface-water Water Quality Index (WQI), and the overall water quality of all reservoirs was classified as “good.” These results indicate that benthic macroinvertebrate eDNA signals in the four reservoirs exhibited significant vertical differences in the composition of layer-specific OTUs and the Simpson diversity index in October, whereas the eDNA-based biological assessment using surface-water samples adequately reflected surface-water quality. Our study provides insights into understanding the vertical distribution of benthic macroinvertebrate eDNA and optimizing biological monitoring strategies in deep reservoirs.