Indicator species models may be a cost-effective approach to estimating species richness across large areas. Obtaining reliable distributional data for indicator species (and therefore reliable estimates of species richness) often requires longitudinal data, that is, surveys for indicator species repeated for several years or time steps. Maximum information must be extracted from such data. We used genetic algorithms and a Bayesian approach to compare the influence of presence/absence data and reporting rate data (the proportion of survey years in which a species was present) on models of species richness based on indicator species. Using data on birds and butterflies from the Great Basin (Nevada, USA), we evaluated models of species richness for one taxonomic group based on indicator species drawn from the same taxonomic group and from a different group. We also evaluated models of combined species richness of both taxonomic groups based on indicator species drawn from either group. We identified suites of species whose occurrence patterns explained as much as 70% of deviance in species richness of a different taxonomic group. Validation tests revealed strong correlations between observed and predicted species richness, with 83–100% of the observed values falling within the 95% credible intervals of the predictions. Whether reporting rate data improved the explanatory and predictive ability of cross-taxonomic models depended on the taxonomic group of the indicator species. The discrepancy in predictive ability was smaller for same-taxon models. Our methods provide a manager with the means to maximize the information obtained from longitudinal survey data.