Large predatory fish that alter the behaviour of smaller species may affect visual estimates of abundance by making organisms more or less difficult to observe and quantify. To evaluate the non-consumptive influence of large predators on rocky reef fishes observed with Baited Remote Underwater Video (BRUV), we tested the hypothesis that fish assemblages monitored when large (ca 1.5 m) piscivorous sharks, Mustelus antarcticus, were present differ from those observed when sharks were absent. We did this in two ways using recordings from rocky reefs in Batemans Marine Park, NSW, Australia. First, we examined 6 min of each 30 min tape; 3 min when sharks were present and 3 min when they were absent, in a paired sample design from 17 sites. Second, we compared fish assemblages for complete tapes (30 min) at sites with sharks present compared to sites where they were absent. The diversity and total abundance of fishes was consistently lower in the presence of sharks; we detected significant assemblage-wide change (PERMANOVA, P < 0.05). Importantly, the diversity and total abundance of fishes for complete tapes (30 min) decreased in the presence of sharks by 18 % and 36 %, respectively, underscoring the likelihood that sharks influence observed fish assemblages. Individual species responses were variable and while the abundance of many fish species decreased in response to sharks, others increased 6-fold. Overall, these results highlight the need to consider systematic differences in the abundance of large predators to avoid biases when testing hypotheses about fish assemblages using non-destructive visual methodologies.