Southern Africa has one of the longest records of fossil hominins and harbours the largest human genetic diversity in the world. Yet, despite its relevance for human origins and spread around the globe, the formation and processes of its gene pool in the past are still largely unknown. Here, we present a time transect of genome-wide sequences from nine individuals recovered from a single site in South Africa, Oakhurst Rockshelter. Spanning the whole Holocene, the ancient DNA of these individuals allows us to reconstruct the demographic trajectories of the indigenous San population and their ancestors during the last 10,000 years. We show that, in contrast to most regions around the world, the population history of southernmost Africa was not characterized by several waves of migration, replacement and admixture but by long-lasting genetic continuity from the early Holocene to the end of the Later Stone Age. Although the advent of pastoralism and farming substantially transformed the gene pool in most parts of southern Africa after 1,300 bp, we demonstrate using allele-frequency and identity-by-descent segment-based methods that the ‡Khomani San and Karretjiemense from South Africa still show direct signs of relatedness to the Oakhurst hunter-gatherers, a pattern obscured by recent, extensive non-Southern African admixture. Yet, some southern San in South Africa still preserve this ancient, Pleistocene-derived genetic signature, extending the period of genetic continuity until today. Oakhurst rockshelter in South Africa documents marked cultural change during the Holocene, but genome-wide data from ancient human individuals at the site now demonstrate a remarkable degree of genetic continuity over the last 9,000 years: the contemporary ‡Khomani San and Karretjiemense from South Africa still show direct signs of relatedness to the Oakhurst hunter-gatherers.

Archaeogenetic research of the last two decades has revealed that the Holocene demographic histories of Stone Age Europe, Asia and North Africa were characterized by several episodes of large-scale migrations, either in the form of admixture with newcomers or by total replacement of the established inhabitants.

In contrast, for South Africa, we demonstrate that the local gene pool was characterized by a prolonged period of genetic continuity with no (detectable) gene flow from outside southern Africa.

The earliest individual in our study that yielded aDNA showed a genetic make-up indistinguishable from the later inhabitants of Oakhurst rockshelter, suggesting that this local ‘southern’ San gene pool was formed more than 10,000 years ago and remained isolated from admixture with neighbouring ‘central’ and ‘northern’ San populations or with more distant sources to the northeast, which admixed with San groups in Malawi and Tanzania.

Our data also demonstrate that subtle fluctuations in the craniofacial size of South African LSA coastal inhabitants (for example, between 4,000 and 3,000 bp) were not the product of genetic discontinuity but probably related to changes in environmental factors or population size. Yet, we highlight that the inhabitants of Oakhurst were not a small, bottlenecked population.

This period of ~9,000 years of genetic continuity ends rather abruptly in the migration events which introduced East and West African-related ancestry to South Africa, accompanied by the spread of herding and farming.