Current spatial patterns of biodiversity are difficult to estimate, but predictions for the future have a set of new challenges. Models are needed that can accurately predict biodiversity for novel environmental conditions, and much focus now is on how best to construct such “transferable” models. Ecological niche models can use occurrence records, the most widely available data for most species, and environmental variables to estimate species ranges, which can then be combined to estimate biodiversity metrics like species richness. Given future estimates of environmental variables, these models can easily make future predictions, but results may be unrealistic for a number of reasons. In this talk, I will discuss some challenges facing future predictions of biodiversity using ecological niche models. Species interactions are not often incorporated into models but can have important effects on broad-scale range patterns. Unchecked model complexity can result in spurious predictions for novel conditions. Finally, taxonomic biases have led to significant knowledge gaps in biodiversity knowledge for invertebrate taxa, which are responsible for essential ecosystem functions and services. I will review these challenges while highlighting relevant research I have done (and plan to do) on these topics, and end with some suggestions on how to move the field of ecological model transferability forward.