Abstract Developing systematic conservation plans depends on a wealth of information on a region's biodiversity. For ‘dark taxa' such as arthropods, such data are usually very incomplete and in most cases left out from assessments. Sky islands are important and often fragile biodiversity hotspots. Southern Appalachian high‐elevation spruce–fir forests represent a particularly threatened sky‐island ecosystem, hosting numerous endemic and threatened species, but their arthropods remain understudied. Here we use voucher‐based megabarcoding to explore genetic differentiation among leaf‐litter arthropod communities of these highlands, and to examine the extent to which they represent dispersed communities of more or less coherent species, manageable as a distributed unit. We assembled a dataset comprising more than 6000 COI sequences representing diverse arthropod groups to assess species richness and sharing across peaks and ranges. Comparisons were standardised across taxa using automated species delimitation, measuring endemism levels by putative species. Species richness was high, with sites hosting from 86 to 199 litter arthropod species (not including mites or myriapods). Community profiles suggest that around one fourth of these species are unique to single sky islands and more than one third of all species are limited to a particular range. Across major taxa, endemicity was lowest in Araneae, and highest in neglected groups like Isopoda, Pseudoscorpionida, Protura and Diplura. Southern Appalachian sky islands of spruce–fir habitat host significantly distinct leaf‐litter arthropod communities, with high levels of local endemicity. This is the first work to provide such a clear picture of peak and range uniqueness for a taxonomically broad sample. Ensuring the protection of a sizeable fraction of high‐elevation litter species richness will therefore require attention at a relatively fine spatial scale.