Environmental classifications are potentially useful tools for summarising broad-scale spatial patterns in ecological and environmental gradients, particularly when biological data are limited in availability. Such classifications can be tuned with respect to specific faunal groups, but their usefulness depends on the validity of the assumption that biological distributions are correlated with gradients in the biophysical environment. Two marine environment classifications with relevance to benthic invertebrate distributions have been developed for New Zealand’s Exclusive Economic Zone (EEZ): the Marine Environment Classification (MEC) and the Benthic Optimised Marine Environment Classification (BOMEC). Until now, however, the ability of the MEC and the BOMEC to map benthic habitats and fauna has not been evaluated against independent sample data. We used benthic invertebrate faunal data from video and epibenthic sled samples collected during Ocean Survey 20/20 (OS 20/20) surveys of Chatham Rise and Challenger Plateau to assess whether the MEC and BOMEC provide a reliable means of mapping benthic habitats and faunal assemblage composition. We also generated a new environmental classification (“Chat/Chall”), tuned solely to the OS 20/20 sample data, to assess the effect of sampling of this type on classification performance. First, we compared the three environmental classifications (MEC, BOMEC, and Chat/Chall) with the OS 20/20 sample sites assigned to a set of 12 biotic habitats which were derived independently by clustering of the faunal data alone. Comparisons were made both visually and using chi-squared tests of goodness-of-fit. Second, using the full multivariate detail of the OS 20/20 faunal data, we used ANOSIM R and homogeneity statistics to assess how well each classification grouped the OS 20/20 sites at all classification levels up to 60 classes. Third, we compared how well each of the environmental classifications grouped the OS 20/20 sample sites in relation to a set of univariate biodiversity metrics calculated for each site from the sample data. None of the three environmental classifications discriminated well between biotic habitats at the site level, but visual comparisons showed consistent patterns which broadly matched distributions of the biotic habitats at larger spatial scales (100–1000 km) in all classifications except the MEC at the 20-class level. These patterns included differentiation between Chatham Rise and Challenger Plateau, and between the north and south flanks of Chatham Rise. ANOSIM R and homogeneity statistic values were low for all three classifications at all class levels, indicating poor ability to map benthic distributions at the spatial scale and taxonomic resolution of the OS 20/20 samples. The classifications also showed poor ability to discriminate the OS 20/20 sites on the basis of biodiversity metrics. The BOMEC and Chat/Chall classifications were generally similar in performance and both were better than the MEC. Our main conclusions from these results are: (1) the BOMEC is an improvement over the MEC for mapping benthic distributions; (2) neither the BOMEC nor the MEC provide reliable information at the spatial scale of individual OS 20/20 sample sites; (3) at larger spatial scales (ca. over 100s km) both MEC and BOMEC classifications produced patterns that were broadly consistent with sampled benthic distributions, suggesting that they might have applications in regional-scale assessment of benthic habitats; (4) to be useful in management or planning applications, objective criteria for determining appropriate, ecologically relevant, classification levels and spatial scales are needed, and (5) further OS 20/20-style surveys could be effective for expanding the scope and generality of existing marine environment classifications.