There has been a rapid rise in papers modelling the impacts of autonomous vehicles. Drawing on a review of this literature, we analyse and discuss the messages conveyed by these studies from a policy-making perspective. An overview of the studies is provided to highlight the different policy frames. We consider the roles that modelling knowledge should have in policy-making in the context of unstable and uncertain futures where novel transport modes bring into question the use of past knowledge and assumptions. We conclude that both limitations of, and assumptions in, models need to be made more explicit and accessible when used in policy-making.