The evolution of opinions in a population of individuals who constantly interact with a common source of user-generated content (i.e. the internet) and are also subject to propaganda is analyzed using computer simulations. The model is based on the bounded confidence approach. In the absence of propaganda, computer simulations show that the online population as a whole is either fragmented, polarized or in perfect harmony on a certain issue or ideology depending on the uncertainty of individuals in accepting opinions not closer to theirs. On applying the model to simulate radicalization, a proportion of the online population, subject to extremist propaganda radicalize depending on their pre-conceived opinions and opinion uncertainty. It is found that an optimal counter propaganda that prevents radicalization is not necessarily centrist.