1Predictive models are frequently used to assist forecasts of species’ distributions. For rare species, this approach can provide insights into habitat utilisation and potential habitat suitability that can assist conservation action. However the technique has seldom been used for invertebrates.2A habitat suitability model was constructed for the rare ant, Formica exsecta, at Glenmore, a pinewood site in Scotland that contains a significant proportion of the UK population of this species. A range of botanical and environmental variables were surveyed within cells in a grid structure. The model successfully re‐classified 87% of cells according to the presence/absence of ant nests. When subject to a jackknife procedure, 70% of cells were successfully re‐classified and the median predicted probability of presence in occupied cells was significantly greater than in unoccupied cells.3The model was further evaluated against three independent datasets with rates for successful predictions of 70%, 62.5% and 25%. The performance declined with increased distance from the area from which calibration data were collected.4We conclude that caution is required when a model is validated purely by reference to the data from which it was constructed and not against independent data. However, the process can provide useful insights into habitat suitability for rare invertebrates that can assist rapid assessment of potential ranges for conservation measures such as restoration of populations.