The unstoppable rise of social networks and the web is facing a serious challenge: identifying the truthfulness of online opinions and reviews. In this paper we use Argumentation Frameworks (AFs) extracted from reviews and explore whether the use of these AFs can improve the performance of machine learning techniques in detecting deceptive behaviour, resulting from users lying in order to mislead readers. The AFs represent how arguments from reviews relate to arguments from other reviews as well as to arguments about the goodness of the items being reviewed.