Preservation of biodiversity is a central goal of conservation management, yet the conditions that promote persistence may differ for the species in the community. For systems subject to stochastic disturbances such as fire, understanding which management practices promote persistence for all species in a community is complex. Before deciding on the best course of action, an objective must be specified. Yet an overarching goal of species persistence can be specified into a measureable objective many different ways. We investigated four alternative management objectives for maximizing species persistence that use common biodiversity indices: (1) attaining the minimally acceptable mix of successional vegetation states to support species' relative abundances, (2) maximizing the arithmetic mean abundance of species, (3) maximizing the geometric mean abundance of species, and (4) minimizing the average extinction risk of species. We used stochastic dynamic programming to model successional changes in vegetation in the presence of both planned and unplanned fires, and utilize an extensive data set on the occurrence of birds, reptiles, and small mammals in different successional states in semiarid Australia. We investigated the influence the choice of objective function and taxonomic focus has on the optimal fire management recommendations. We also evaluated a recent hazard reduction policy to annually burn a fixed amount of the landscape and compare results to the optimal solution. The optimal management strategy to maximize species persistence over a 100-year period is predominantly to minimize wildfires. This is because the majority of species are more likely to occur in intermediate and late successional vegetation. However the optimal solution showed sensitivity to the objective and the species included in the analysis. These results highlight the need for careful consideration when specifying an objective to represent overarching conservation goals. Using the extinction risk objective, we show that a policy to annually burn 5% of the landscape could increase the average probability of extinction for the modelled species by 7% over the next 100 years compared to the optimal management scenario.