A method for controlling distribution of media contents over a network, wherein the contents are distributed by making the contents available at surrogate servers. The method includes the steps of identifying additional contents eligible for distribution; defining a set of categories; identifying for each category at least a reference content; associating the additional contents with the predefined categories based on semantics affinity with the reference content, the semantics affinity being calculated as the distance of each of the additional contents to the at least a reference content; selecting at least one of said predefined categories; and making at least one of the additional contents associated with said selected predefined category available for distribution at the surrogate servers.