Arthropods are essential for maintaining soil ecosystem functioning. However, their full diversity remains insufficiently explored due to their taxonomic complexity, the limitations of traditional methods such as morphotaxonomy, and constraints associated with DNA barcoding approaches. In this study, we applied COI-based DNA metabarcoding to survey soil arthropod richness and community composition at a single timepoint across 28 long-term monitoring sites in Switzerland representing three land-use types, i.e., arable land, permanent grassland, and forest. A total of 6464 arthropod amplicon sequence variants (ASVs) were identified, with forests exhibiting the highest richness, followed by permanent grasslands and arable land. Key environmental drivers of soil arthropod community composition included soil C:N ratio, mean annual temperature, and metabolic quotient. The arthropod ASVs were classified into 389 distinct taxa, including 214 species-level assignments. The maximum number of ASVs assigned to a single species was 180 for the collembolan Lepidocyrtus cyaneus. More detailed phylogenetic analyses of the genus Lepidocyrtus, containing 232 ASVs assigned to 4 species, revealed that i) ASVs assigned to a species and reference sequences annotated as that species were often distributed across several phylogenetic clusters, whereas in other cases phylogenetic clusters included sequences with different species-level assignments or taxonomic annotations, and ii) phylogenetic clusters were associated with specific land-use types or land-use type combinations, indicating potential hidden diversity. Two additional parthenogenetic collembolan species, Folsomia candida and Parisotoma notabilis, included more than 100 ASVs, revealing a high potential for hidden diversity. Together, our results highlight the potential of arthropod-specific metabarcoding combined with phylogenetic analyses to characterize hidden diversity in complex mesofaunal communities by linking ecological and phylogenetic information.