The marine environment offers few obvious barriers to dispersal for broadcast‐spawning species, yet population genetic structure can occur on a scale much smaller than the theoretical limits of larval dispersal. Comparative phylogeographical studies of sympatric sister species can illuminate how differences in life history, behaviour, and habitat affinity influence population partitioning. Here we use a mitochondrial DNA marker (612 bp of cytochrome c oxidase subunit I) to investigate population structure of three endemic Hawaiian broadcast‐spawning limpets (Cellana spp.) with planktonic larvae that are competent to settle within 4 days. All three species exhibit significant population structure and isolation by distance, but the spatial scales of partitioning differ among the species. Cellana talcosa (n = 105) exhibits strong population structure between Kauai and the other main Hawaiian Islands (MHI) where the maximum channel width is 117 km, and no shared haplotypes were observed (Φ_CT = 0.30, P < 0.001). In contrast, populations of Cellana exarata (n = 149) and Cellana sandwicensis (n = 109) exhibit weaker population structure within the MHI (Φ_ST = 0.03–0.04, P < 0.05), and between the MHI and the Northwestern Hawaiian Islands (Φ_ST = 0.03–0.09, P < 0.01), where the maximum channel width is 260 km. Biogeographical range and microhabitat use were correlated with estimates of dispersal, while phylogenetic affiliation and minimum pelagic larval duration were poor predictors of population partitioning. Despite similar life histories, these closely related limpets have contrasting patterns of population structure, illustrating the danger of relying on model species in management initiatives to predict population structure and dispersal in the context of marine protected area delineation.