The modern directions of development of fan fiction as a media system that has acquired the characteristics of self-organization are actualized. Three stages of development of scientific views on the formation of the process of collective authorship are presented: from narrative criticism and isolation of media features of fandoms to comprehension of the facts of the reverse influence of fan fiction on culture and communication processes. On the example of the development of modern fan fiction space, the phenomena that express the communicative nature of the fan fiction community, as well as the network nature of its organization are proposed for analysis: creation of podcast systems; according to the logic of canonical blockbuster universes, the development of complex multi-story stories with the involvement of a large number of participants. The paradoxical phenomena that arise in this media system - the growth of original works not related to rethinking and implementing alternative lines of the canon, the emergence of the phenomenon of the passive spectator - are explained by the influence of general cultural trends and local national practices. Indirectly, the influence of fan-fiction activity is presented in various manifestations and trends: the phenomena of secondary and tertiary communication, the transformation of the fan into canonical texts for further fiction, the active departure of fans outside closed communities and the impact on modern cultural practices. Examples of the latter are illustrated by the use of fan fiction in writing scripts for series; creation of spin-products; taking into account fan thoughts when developing the plots of TV series and books. The prospects of communication research are motivated by the dynamics of growth of the object of study and the complexity of modern methods of analysis. The presence of contradictory tendencies and manifestations problematizes the finality and immutability of already established scientific approaches, forces to turn to technical approaches using big-date methods.