Abstract Many ecological functions, including seed dispersal, pollination and the flow of nutrients between ecosystems, result from the movement of animals. The ongoing reduction and extinction of animal populations and species means that these functions are reduced or lost. As a result, conservation and restoration activities often seek to reinstate ecological functions to the communities from which they have been depleted. Advances in movement and allometric ecology offer a way to reconstruct the movement and emergent functions of extinct taxa. Here, we describe a framework for integrating insights from movement ecology with allometric relationships linking body mass to movement and other traits and demonstrate how this can be used to reconstruct the emergent functions of extinct taxa. Allometric relationships are best resolved for vertebrates (especially mammals), so we highlight potential gaps relevant to developing trait‐based allometric movement models for data‐depauperate taxa. To test our framework, we use it to estimate seed dispersal kernels for seven extant bird species for which seed dispersal distances are available. We then apply it to predict the movement activity and emergent seed dispersal kernels for three extinct endemic New Zealand bird species (two flightless and one volant), spanning four orders of magnitude in body mass. The model performs satisfactorily for the extant species, with predicted mean, median and maximum seed dispersal distances overlapping empirical estimates, but with a tendency to over‐prediction. If empirical data were available, the model could be calibrated for specific taxa or landscapes. The model shows that the three extinct bird species could have been important agents of long‐distance seed dispersal, and comparable functions are probably now absent in New Zealand's ecosystems. Summary. Trait‐based allometric models can be used to simulate the movement and associated functions of extinct animals. Reconstructing the functions of extinct (or data depauperate) species cannot be achieved by chaining together statistical relationships; such efforts must adequately represent natural history and carefully use computational model evaluation tools. A further challenge is adequately representing the effects of past landscape heterogeneity on disperser movement, fruit distribution and hence dispersal distances. A free Plain Language Summary can be found within the Supporting Information of this article.