Abiotic and biologically informed classifications are often used in conservation planning as coarse-filter surrogates for species. The relationship between these surrogates and the distribution of species is commonly assumed, but rarely assessed by planners. We derived four abiotic and eight biologically informed classifications of stream reaches to serve as surrogates for biodiversity patterns in the Wet Tropics bioregion, Queensland, Australia. We used stream reaches as planning units and, as conservation targets for each surrogate, we used two percentages – 10% and 30% – of the total stream reach length occupied by each class. We then derived minimum sets of planning units to meet targets for each surrogate and tested the effectiveness of the surrogates by calculating the average achievement of the same targets for predicted distributions of 28 fish species. Our results showed that neither abiotic nor biologically informed classifications were good at representing freshwater fish species; in fact none of the surrogates led to average representation of species better than randomly selected planning units. There were two main reasons for this poor performance. First, none of the surrogates had high classification strength or informativeness about compositional change in fish species within the study region. Second, frequency distributions of probabilities of occurrence for most fish species were strongly right-skewed, with few stream reaches having high probabilities. Combined, these results meant that selection of stream reaches to achieve surrogate targets was effectively random with respect to probabilities of fish species occurrence, leading to poor representation of fish species. We conclude there is a limited basis for using coarse-filter surrogates to represent freshwater fish diversity in this region, and that there is a clear need for research in this as well as other regions if planners are to understand the limitations associated with coarse-filter surrogates for representing freshwater biodiversity more broadly.