Fences are increasingly used globally as a management tool in conservation to reduce wildlife depredations, disease transmission, and wildlife mortality. There are a limited number of studies on the genetic effects of perimeter fencing of protected areas on megaherbivores. Using population genetic analyses on 226 sequences of a 400 bp fragment of the mtDNA Dloop from 10 East African buffalo populations (3 fenced and 7 unfenced), the influence of spatial isolation and fencing on buffalo population genetic diversity and genetic differentiation were examined. Mean gene diversity between fenced and unfenced buffalo populations was not different (fenced: 0.978 ± 0.003, unfenced: 0.973 ± 0.004, P = 0.300) but nucleotide diversity was higher in fenced than unfenced populations (fenced: 0.038 ± 0.019, unfenced: 0.030 ± 0.015, P = 0.005). Genetic differentiation among buffalo populations based on haplotype frequencies and model-based genetic distance was weak (FST = 0.08, ΦST = 0.06) and contributed to 6.2% and 8.5% of total genetic variance respectively. Ninety three percent of population pairs were genetically differentiated by distances determined from haplotype frequencies but only 51% of population pairs were differentiated using modelled distances suggesting recent differentiation. There was no correlation between linearized FST and geographical distance (r = -0.005, P = 0.52), but linearized ΦST was moderately correlated with geographic distance (r = 0.329, P = 0.03). The distance effect was greater when fenced populations were excluded (ΦST: r = 0.464, P = 0.05), suggesting that insularization due to fencing is distorting isolation by distance. SSD analyses revealed that 2 of 3 fenced populations and 2 of 7 unfenced populations had non-unimodal distributions suggesting demographically declining populations. Our study reveals the high genetic diversity but warn that genetic erosion due to isolation, including fencing, is only beginning to have an impact on East African buffalo populations.