The diversity and community structure of invertebrates in the water and sediment of Lake Haizhu in the Guangzhou Haizhu National Wetland Park was investigated based on environmental DNA metabarcoding technology (eDNA), and the detection ability for zooplankton was compared between eDNA and traditional morphological identification. The results showed that a total of 9 phyla, 16 classes, 34 orders, 71 families, 93 genera and 137 operational taxonomic units (OTUs) of aquatic invertebrates were detected by eDNA. In the samples directly collected from water (W), a total of 9 phyla, 50 genera and 68 OTUs were detected; in the samples collected by a planktonic net (N), a total of 6 phyla, 27 genera and 35 OTUs were detected; and in the sedimental samples (S), a total of 9 phyla, 70 genera and 103 OTUs were detected. Rotifera and Arthropoda were the major invertebrates in both W and N; Annelida and Arthropoda were the major invertebrates in S. The OTU number was highest in S, followed by W and N. The copepods (100%) and most of rotifers (58.82%) identified by the morphological method could be annotated by eDNA, while most of the copepods (71.43%) and rotifers (58.82%) annotated by eDNA were not detected by the morphological identification; and cladocerans was not annotated by eDNA. The results suggest that eDNA has high potential for invertebrate surveys, and the combination of eDNA and traditional morphological identification provides a more comprehensive understanding of invertebrates and aquatic biodiversity of the lake ecosystem.