Species richness is not evenly distributed across the tree of life and a limited number of lineages comprise an extraordinarily large number of species. In lichen-forming fungi, only two genera are known to be ‘ultradiverse’ (>500 species), with the most diverse genus, Xanthoparmelia, consisting of c. 820 species. While Australia and South Africa are known as current centres of diversity for Xanthoparmelia, it is not well known when and where this massive diversity arose. To better understand the geographical and temporal context of diversification in this diverse genus, we sampled 191 Xanthoparmelia specimens representing c. 124 species/species-level lineages from populations worldwide. From these specimens, we generated a multi-locus sequence data set using Sanger and high-throughput sequencing to reconstruct evolutionary relationships in Xanthoparmelia, estimate divergence times and reconstruct biogeographical histories in a maximum likelihood and Bayesian framework. This study corroborated the phylogenetic placement of several morphologically or chemically diverse taxa within Xanthoparmelia, such as Almbornia, Chondropsis, Karoowia, Namakwa, Neofuscelia, Omphalodiella, Paraparmelia, Placoparmelia and Xanthomaculina, in addition to improved phylogenetic resolution and reconstruction of previously unsampled lineages within Xanthoparmelia. Our data indicate that Xanthoparmelia most likely originated in Africa during the early Miocene, coinciding with global aridification and development of open habitats. Reconstructed biogeographical histories of Xanthoparmelia reveal diversification restricted to continents with infrequent intercontinental exchange by long-distance dispersal. While likely mechanisms by which Xanthoparmelia obtained strikingly high levels of species richness in Australia and South Africa remain uncertain, this study provides a framework for ongoing research into diverse lineages of lichen-forming fungi. Finally, our study highlights a novel approach for generating locus-specific molecular sequence data sets from high throughput metagenomic reads.