BackgroundThe cox1-barcoding approach is currently extensively used for high-throughput species delimitation and discovery. However, this method has several limitations, particularly when organisms have large effective population sizes. Paradoxically, most common, abundant, and widely distributed species may be misclassified by this technique.ResultsWe conducted species delimitation analyses for two host-specific lineages of scab mites of the genus Caparinia, having small population sizes. Cox1 divergence between these lineages was high (7.4–7.8%) while that of nuclear genes was low (0.06–0.53%). This system was contrasted with the medically important American house dust mite, Dermatophagoides farinae, a globally distributed species with very large population size. This species has two distinct, sympatric cox1 lineages with 4.2% divergence. We tested several species delimitation algorithms PTP, GMYC, ABGD, BPP, STACEY and PHRAPL, which inferred different species boundaries for these entities. Notably, STACEY recovered the Caparinia lineages as two species and D. farinae as a single species. BPP agreed with these results when the prior on ancestral effective population sizes was set to expected values, although delimitation of Caparinia was still equivocal. No other cox1 species delimitation algorithms inferred D. farinae as a single species, despite the fact that the nuclear CPW2 gene shows some evidence for introgression between the cox1 groups. This indicates that the cox1-barcoding approach may result in excessive species splitting.ConclusionsOur research highlights the importance of using nuclear genes and demographic characteristics to infer species boundaries rather than relying on a single-gene barcoding approach, particularly for putative species having large effective population sizes.Electronic supplementary materialThe online version of this article (10.1186/s13071-018-3242-5) contains supplementary material, which is available to authorized users.