The pervasive deployment of P2P (peer-to-peer) systems and the multimedia contents overload in web environment raise a serious complexity for the peers where peers that participate in a P2P network are no longer able to effectively choose the contents they want. Recommender systems have been popularly used for reducing information overload of internet surfers by suggesting products or digital contents that are most valuable for them. But most existing recommender systems have been worked in client–server architecture. This paper proposes a PEOR (PEer-ORiented Recommender system), a collaborative filtering-based multimedia contents recommender system in P2P architecture, to obtain the peers’ search efficiency. To adopt a change in peer preferences PEOR uses recent ratings of peers for recommendations, thereby leading to better quality recommendations. And to enhance the system performance, PEOR searches for nearest peers with similar preference through peer-based local information only. We implemented the system and evaluated its performance with real transaction data in S content provider offering character images. Our experimental data shows that PEOR offers not only remarkably higher quality of recommendations but also the dramatically faster performance than the centralized benchmark system.