The risk of collision between birds and turbines is seen as one of the key issues in the planning process for offshore wind farms. In some cases, predictions of collision risk have led to projects either being withdrawn from the planning process, or refused planning consent. Despite this, the evidence base on which collision risk is assessed is extremely limited and assessments rely on models which can be highly sensitive to assumptions, notably about bird collision avoidance behaviour. We present a synthesis of the current state of knowledge about collision risk and avoidance behaviour in seabirds. Evidence suggests species-specific responses to turbines and that in order to avoid collision, most birds adjust their flight paths at some distance from the turbines, rather than making last-second adjustments. We highlight the key gaps in knowledge and make recommendations for future data collection.