This paper studies a model for describing opinions’ evolution in a community linked by microblogging directed network systems. The network can evolve in time by means of the creation and the deletion of some connections. When a connexion is created, the individuals have access to the whole history of posts written by the corresponding author and, when a connexion is destroyed, all the posts written by the corresponding author become invisible. The agents’ opinions are described by a set of continuous opinion variables in the closed interval [Formula: see text]. They represent the agreement (or disagreement) of the corresponding agent with respect to a binary question (such as a referendum or an election with two candidates). The model takes into account the effects on public opinion caused by the sign and the intensity of the initial opinions of the agents, their activity in microblogging platforms the presence of leaders and the possible manipulations of the visibility of the posts by the microblogging platform owner. The model is written as a system of integro-differential equations and simulated by using a Runge–Kutta method. We show that hidden manipulation can have an important impact on the public opinion formation and that very mild interventions of the network owner may induce major effects on the population. In our simulations, the effect of hidden manipulation is shown to be more efficient in driving the public opinion than the action of a leader, in the case of bounded confidence models and for short-time intervals.