AbstractMitochondrial DNA (mtDNA) has long been used to date the divergence between species, and to explore the time when species’ effective population sizes changed. The idea that mitochondrial DNA is useful for molecular dating rests on the premise that its evolution is neutral. This premise was questionable to begin with, and even though it has long been challenged, the evidence against clock-like evolution of mtDNA is usually ignored. Here, we present a particularly clear and simple example to illustrate the implications of violations of the assumption of selective neutrality. DNA sequences were generated for the mtDNA COI gene and the nuclear 28S rRNA of two closely related and widely distributed rocky shore snails whose geographical ranges are defined by different thermal preferences. To our knowledge, this is the first study to use nuclear rRNA sequence for studying species-level genealogies instead of phylogenetics, presumably because this marker is considered to be uninformative at this taxonomic level. Even though the COI gene evolves at least an order of magnitude faster, which was reflected in high inter-specific divergence, intraspecific genetic variation was similar for both markers. As a result, estimates of population expansion times based on mismatch distributions were completely different for the two markers. Assuming that 28S evolves effectively clock-like, these findings likely illustrate variation-reducing purifying selection in mtDNA at the species level, and an elevated divergence rate caused by divergent selection between the two species. Although these two selective forces together make mtDNA suitable as a DNA barcoding marker because they create a ‘barcoding gap’, estimates of demographic change can be expected to be highly unreliable. Our study contributes to the growing evidence that the utility of mtDNA beyond DNA barcoding is limited.