An approach to downloading content to a mobile device based on predicted service disruptions to avoid a loss of access to content. The service disruptions to wireless network connectivity can occur due to network dead spots, such as those occurring on a highway, in a tunnel, under a bridge or in a building. Embodiments monitor current and historical mobile device usage and historical wireless service data to form a history and generate a predictive model that makes predictions regarding anticipated dead spots and the types and quantity of content that will be requested during the dead spots. Decisions based on the predictive model can be made automatically by embodiments, such as when to start downloading content prior to entering a dead spot and how much content and what types of content will be downloaded based on predicted user requests.