Ancient hybridization is difficult to detect and is often surmised from conflicting patterns between phenotypic and genetic data sets that are difficult to explain with alternative hypotheses. Here, a fortuitous macromutation in a microsatellite locus allows an unusually distinctive footprint of ancient introgression to be inferred between two highly divergent bird species that began to speciate in the Late Miocene. A cline of distinctive high-repeat-number (large) alleles at a locus in the Mexican Jay (Aphelocoma wollweberi), which predominantly has low-repeat-number (small) alleles, reaches its highest frequency near the range border with Transvolcanic Jays (A. ultramarina), which are fixed for large alleles, indicating gene flow. The gene flow is not recent, however, because very few of the allelic states are shared between Mexican Jays and Transvolcanic Jays. Ancient gene flow is also more plausible than current gene flow because prior mitochondrial DNA results show no evidence of current dispersal between the ranges of the species, and Mexican Jays with introgressed large alleles are not phenotypically more similar to Transvolcanic Jays, as would be expected with recent hybridization. Although bifurcations in the tree of life are often assumed to belie complex histories of introgression, our results provide some of the clearest evidence that speciation can survive one or multiple bouts of gene flow with no detectable trace in the phenotypes and mtDNA of the constituent species.