Biogeography of Sideroxylon (Sapotoideae) and whether the satellite genera Argania and Spiniluma merit recognition are revisited. The hypothesis of an African origin with a subsequent migration to Central America via Europe and the North Atlantic landbridge is challenged. We analysed 58 accessions of trnH-psbA and ITS sequences in a fossil-calibrated, relaxed lognormal clock model with BEAST for phylogenetic and biogeographic inference. Argania spinosa (L.) Skeels from Morocco must be united with Sideroxylon, whereas S. discolor Radcl.-Sm. and S. oxyacanthum Baill. belong to subfamily Chrysophylloideae and may be recognised as Spiniluma. The divergence time estimate suggests that Sideroxylon originated and first diversified in Central America 56.3–52.2 million years ago, contemporaneous with the landbridge, but an archipelago of islands and the Tethys Seaway halted eastward expansion to Africa until c. 20 million years ago, i.e. 25 million years after establishment in Africa. Range expansion of Sideroxylon was therefore not powered by the landbridge, and a long-distance dispersal from Central America to Africa is proposed. The establishment of the Gomphotherium landbridge between Africa and Eurasia at 19 million years ago provided a land-migration route to Europe and Asia, which is reconcilable with the extension of Xantolis into Asia. Sideroxylon has colonised Socotra by over-water dispersal in the Gulf of Aden, Macaronesia, Madagascar and the Mascarene Islands.