We propose a model of product reviews in which some are genuine and some are fake in order to shed light on the value of information provided on platforms like TripAdvisor, Yelp, etc. In every period, a review is posted which is either genuine or fake. We characterize the equilibrium of the dynamic model and prove that it is unique. In equilibrium, valuable learning takes place in every period. Fake reviews, however, do slow down the learning process. It is established that any attempt by the platform to manipulate the reviews is counterproductive.