Environmental DNA (eDNA) offers valuable presence/absence data for populations and has been widely used in comprehensive biodiversity assessments. However, applying eDNA in terrestrial environments poses unique challenges, particularly in obtaining samples that are representative of ecological communities. eDNA extracted from top-predator dietary samples can be an effective sampling source in monitoring prey populations. In this study, we tested a novel, non-destructive protocol to assess the efficacy of eDNA from barn owl (Tyto javanica delicatula) pellets as a tool for monitoring small mammal communities in an arid environment. We assessed the species composition and abundance of small mammals from owl pellets collected in the Simpson Desert in far western Queensland, Australia, using a three-tiered approach. We extracted DNA from 50 owl pellets and targeted a 16S mini-barcode for metabarcoding. We compared species detection via genetic analysis with that of morphological analysis, and finally with historical small mammal trapping data. The DNA extraction method presented here resulted in full preservation of prey bones and fur material for museum archival. eDNA detected four mammal species that were not detected via morphological pellet analysis, three of which are significant detections that had not been observed at this location before but were expected to occur based on likely distribution ranges. However, a key limitation of the eDNA approach demonstrated in this study, is that taxonomic identification was constrained by the completeness of reference databases, which can result in false negatives or ambiguous assignments. The results of the present study demonstrate that the specificity of an eDNA approach can offer advantages compared with morphological identification of mammalian remains from owl pellets, and that genetic owl pellet analysis may be particularly useful in full vertebrate diversity assessments that include reptiles, birds and amphibians that are unidentifiable from skeletal remains.