We compared and contrasted fish assemblage data sampled by baited remote underwater stereo-video systems (stereo-BRUVs) and stereo-video remotely operated vehicles (stereo-ROVs) from subsea pipelines, reef and soft sediment habitats. Stereo-BRUVs sampled greater fish diversity across all three habitats, with the stereo-ROV sampling ~46% of the same species on pipeline and reef habitats. Larger differences existed in soft sediment habitats, with stereo-BRUVs recording ~65% more species than the stereo-ROV, the majority of which were generalist carnivores. These differences were likely due to the bait used with stereo-BRUVs attracting fish from a large and unknown area. Fish may have also avoided the moving stereo-ROV, an effect possibly magnified in open soft sediment habitats. As a result of these biases, we recommend stereo-ROVs for assessing fish communities on pipelines due to their ability to capture fish in-situ and within a defined sampling area, but caution is needed over soft sediment habitats for ecological comparisons. •Stereo-BRUVs and Stereo-ROV sampled different compositions of fish on and off the pipeline.•Stereo-ROV may under-sample fish in open sandy habitats.•Stereo-BRUVs may over-sample fine-scale habitat associations due to bait attraction.•Stereo-ROV samples fish-habitat associations in greater detail.•Stereo-ROV is a more suitable tool for sampling fish on pipeline and fine-scale structures.