Discriminating marine fish assemblages at broad scales can be difficult because of heterogeneity within their habitats, variability in patterns of behaviour and abundance of fish between habitats, and sampling biases in extractive fishing techniques when used across a range of habitats. Remote underwater video stations have recently been developed to help overcome these problems, but the use of bait as an attractant raises questions about bias towards scavengers and predators in samples of fish communities. We compared the ability of baited and unbaited underwater video stations to discriminate between fish assemblages inhabiting distinct benthic habitats in temperate and tropical continental shelf waters in Australia, to help test whether the bait attracted predatory and scavenging species to the video in disproportionate numbers in comparison to other trophic groups, such as herbivores. Data from baited video cameras displayed a clearer discrimination in constrained canonical analysis of principal coordinates of fish assemblages between marine habitats in both tropical and temperate environments. Analysis of the key trophic groups indicated that bait attracted greater numbers of predatory and scavenging species without decreasing the abundances of herbivorous or omnivorous fishes. There was greater similarity between replicate samples from baited video within habitats, implying that the use of bait will provide better statistical power to detect spatial and temporal changes in the structure of fish assemblages and the relative abundances of individual species within them.