Social network analysis provides valuable opportunities to quantify the nature of social relationships in animal societies including aspects of group structure, dynamics and behaviour transmission. Remote monitoring approaches such as camera trapping offer rich data sets from groups and species that are difficult to observe, yet the robustness of these data for constructing social networks remains unexplored. Here we compared networks of party association based on camera traps with those based on direct observations over the same 9-month sampling period in a group of habituated western chimpanzees, Pan troglodytes verus. Networks based on camera traps and direct observations were both stable with sufficient sampling, and had very similar structures, patterns of sex assortment and individual network positions. However, camera trap data led to lower estimates of group density and dyadic association strengths, and slightly higher modularity, illustrating the limitations raised by differences in data collection methods for network comparisons. We then constructed a social network using camera trap data from unhabituated eastern chimpanzees, P.t. schweinfurthii, demonstrating the feasibility of this approach in the absence of extensive prior knowledge of the study subjects. Further, differences between the eastern and western chimpanzee social networks followed expected patterns based on recognized social differences, illustrating the promise of this approach for detecting within-species social variation. Although long-term behavioural observations will continue to provide rich data for many species, camera traps offer a powerful alternative to gain information on social group dynamics in elusive or unhabituated animals, as well as to conduct systematic multisite comparative studies.