The philopatric behaviour of albatrosses has intrigued biologists due to the high mobility of these seabirds. It is unknown how albatrosses maintain a system of fragmented populations without frequent dispersal movements, in spite of the long-term temporal heterogeneity in resource distribution at sea. We used both genetic (amplified fragment length polymorphism) and capture–mark–recapture (CMR) data to identify explicitly which among several models of population dynamics best applies to the wandering albatross (Diomedea exulans) and to test for migration–drift equilibrium. We previously documented an extremely low genetic diversity in this species. Here, we show that populations exhibit little genetic differentiation across the species’ range (ΘB < 0.05, where ΘB is an FST analogue). Furthermore, there was no evidence of hierarchical structure or isolation-by-distance. Wright's FST between pairs of colonies were low in general and the pattern was consistent with a nonequilibrium genetic model. In contrast, CMR data collected over the last decades indicated that about one bird per cohort has dispersed among islands. Overall, FST values were not indicative of contemporary dispersal as inferred from CMR data. Moreover, all genotypes grouped together in a cluster analysis, indicating that current colonies may have derived from one ancestral source that had a low genetic diversity. A metapopulation dynamics model including a recent (postglacial) colonization of several islands seems consistent with both the very low levels of genetic diversity and structure within the wandering albatross. Yet, our data suggest that several other factors including ongoing gene flow, recurrent long-distance dispersal and source-sink dynamics have contributed to different extent in shaping the genetic signature observed in this species. Our results show that an absence of genetic structuring may in itself reveal little about the true population dynamics in seabirds, but can provide insights into important processes when a comparison with other information, such as demographic data, is possible.