In order to accurately assess the status of the cheetah Acinonyx jubatus it is necessary to obtain data on numbers and demographic trends. However, cheetahs are notoriously difficult to survey because they occur at very low population densities and are often shy and elusive. In South Africa the problem is further complicated in areas where land is privately owned, restricting access, with dense bush and cheetahs that are frequently persecuted. Cheetahs are individually identifiable by their unique spot patterns, making them ideal candidates for capture–recapture surveys. Photographs of cheetahs were obtained using four camera traps placed successively at a total of 12 trap locations in areas of known cheetah activity within a 300 km² area in the Thabazimbi district of the Limpopo Province. During 10 trapping periods, five different cheetahs were photographed. These results were used to generate capture histories for each cheetah and the data were analysed using the capture–recapture software package CAPTURE. Closure tests indicated that the population was closed (P = 0.056). The M_h model was used to deal with possible heterogeneous capture probabilities among individual cheetahs. Closure tests did not reject the model assumption of population closure (P = 0.056). The M_h model produced a capture probability of 0.17 with an estimate of 6–14 cheetahs (P = 0.95) and a mean population size of seven cheetahs (S.E. = 1.93). These results are promising and will be improved with employment of more camera traps and sampling a larger area.