Abstract: Systems and methods for predictively identifying content sources as likely activity subscribers to an online service is presented. In view of a corpus of activity data describing activity of users of an online service, an analysis of this corpus of activity data is carried out to identify one or more content sources that are not activity subscribers but whose content is posted on the online service. A machine learning model is employed to predictively identify at least some content sources that are not activity subscribers of the online service that could likely become activity subscribers. A first content source of the one or more content sources is identified as a likely potential activity subscriber according to the trained machine learning model and the first content source is notified with information for becoming an activity subscriber.