Graph clusters are good for recommender systems. These systems try to find similarity among users and therefore cluster them. Once clustered they can find new topic from other nodes in the same cluster and recommend them to the nodes which have not seen those topics yet.