As software applications become increasingly sophisticated and complex, it is harder to ensure speed and responsiveness. For example, a sizeable fraction of content displayed by mobile applications is fetched in real time from the internet, because, e.g., it is not already available on the user device. To improve speed and responsiveness, applications attempt to anticipate user action and pre-fetch content. Techniques of this disclosure determine the screen area where a user is focusing by using gaze detection techniques, for users that provide permission for such detection. The gaze detection is performed using a front facing sensor of a user device and a model, e.g., a trained machine-learning model, or heuristics. With user permission, detected gaze data is provided to applications that can utilize such data to prefetch content, thereby optimizing usage of the data network and allowing for faster app interaction. Gaze detection can be implemented in an operating system, or within an application.