Online Social Networks (OSNs) have become immensely popular for spammers spreading malicious content and links. Using the nature of OSNs, spammers frequently change their behavior to avoid detection. The current approaches for spam detection are mainly based on classification techniques. Classification techniques have known issues such as “data labelling”, “spam drift”, “imbalanced datasets” and “data fabrication” which hinder the performance and the accuracy of spam detection. In this paper we propose a fully unsupervised approach using a user's peer acceptance within the OSN to distinguish spammers from legitimate users. Peer acceptance can be derived based on common shared interests over multiple shared topics. The contribution of this research is an unsupervised method to detect spammers based on users’ peer acceptance generated from users’ post content. While not as accurate as traditional supervised classification techniques, our unsupervised techniques are able to achieve 94.1% accuracy on some datasets without the need for labelling.