Baited remote underwater video stations (BRUVS) are increasingly being used to examine assemblages of fishes, yet critical methodological questions related to sampling limitations and bias, such as the influence of bait type, remain poorly understood. At multiple locations, we examined the hypothesis that diversity and abundance in temperate reef fish assemblages were independent of bait type. We used 3 bait types (abalone viscera, pilchards and crushed urchin) and quantified commonly used metrics for the fish assemblage, including species richness, time of first arrival and relative abundance on 3 shallow rocky reefs in southeastern Australia over 2 yr. We distinguished the following 6 feeding guilds: herbivore, zooplanktivore, alga/invertebrate consumers, invertebrate carnivore, macroinvertebrate carnivore and generalist carnivore. The response of fishes was dependent on bait type, with urchin bait performing particularly poorly. Although we did not detect statistical differences between the performance of pilchards and abalone viscera as bait, pilchards produced more consistent outcomes. Importantly, we also observed strong spatial effects. In general, bait type had a marked effect on species richness, but little influence on relative abundance. Overall we conclude that oily bait such as pilchards, which have been widely used in most studies, yield the most consistent outcomes. Consequently, bait type and spatial variation in fish assemblages needs to be considered in sampling designs to assess the limitations of BRUVS.