Understanding recent speciation history requires merging phylogenetic and population genetics approaches, taking into account the persistence of ancestral polymorphism and possible introgression. The emergence of a clear phylogeny of hares (genus Lepus) has been hampered by poor genomic sampling and possible occurrence of mitochondrial DNA (mtDNA) introgression from the arctic/boreal Lepus timidus into several European temperate and possibly American boreal species. However, no formal test of introgression, taking also incomplete lineage sorting into account, has been done. Here, to clarify the yet poorly resolved species phylogeny of hares and test hypotheses of mtDNA introgression, we sequenced 14 nuclear DNA and 2 mtDNA fragments (8205 and 1113 bp, respectively) in 50 specimens from 11 hare species from Eurasia, North America, and Africa. By applying an isolation-with-migration model to the nuclear data on subsets of species, we find evidence for very limited gene flow from L. timidus into most temperate European species, and not into the American boreal ones. Using a multilocus coalescent-based method, we infer the species phylogeny, which we find highly incongruent with mtDNA phylogeny using parametric bootstrap. Simulations of mtDNA evolution under the speciation history inferred from nuclear genes did not support the hypothesis of mtDNA introgression from L. timidus into the American L. townsendii but did suggest introgression from L. timidus into 4 temperate European species. One such event likely resulted in the complete replacement of the aboriginal mtDNA of L. castroviejoi and of its sister species L. corsicanus. It is remarkable that mtDNA introgression in hares is frequent, extensive, and always from the same donor arctic species. We discuss possible explanations for the phenomenon in relation to the dynamics of range expansions and species replacements during the climatic oscillations of the Pleistocene.