Aim Underpinning conservation and management strategies at large spatial scales is the concept that spatial patterns of biodiversity are known, although this information is frequently lacking. Many countries routinely collect data on fish occurrence as part of stock assessments, and it has been suggested that this data could be a surrogate for other components. Here we test the usefulness of three indices of demersal fish species diversity as a surrogate for epibenthic richness at scales from 0.002 to 3,800,000 km À2 .Location Four areas covering a spatial extent of 11°of latitude and 18°of longitude within the New Zealand Territorial Sea and Exclusive Economic Zone.Methods Predictive modelling of epibenthic taxon richness, from video and sled-based dredge, by boosted regression trees using environmental variables and demersal fish taxon richness.Results Overall, the ability of environmental factors to predict epibenthic richness was not significantly improved by incorporating any of the demersal fish richness estimates, either within or across locations. However, incorporating data on fish richness collected at the same time and spatial scale (c. 2 km) as the video data demonstrated a small increase (c. 10%) in predictability. The ability of environmental variables to predict epibenthic richness varied within and across locations, from 32% to 72% variability explained, despite the presence of strong environmental gradients.Main conclusion Demersal fish taxon richness, especially derived from stock assessments, was not a useful surrogate for epibenthic invertebrate taxon richness. Environmental data were more useful in predicting epibenthic richness, especially for habitat-structuring taxa observed on video. Our findings suggest that the need to manage large areas of the seafloor seems most likely to be met in future by species distribution and habitat suitability mapping. However, understanding small-scale spatial heterogeneity may be important for conservation and management strategies operating at a range of scales.