Automatic face annotation (or tagging) facilitates improved retrieval and organization of personal photos in online social networks. In this paper, we present a new collaborative face recognition (FR) method that aims to improve face annotation accuracy. The proposed method makes efficient use of multiple FR engines and databases that are distributed over an online social network. The performance of our collaborative face recognition method was successfully evaluated using the standard MPEG-7 VCE-3 data set and a set of real-world personal photos from the Web. The efficacy of the proposed method is demonstrated in terms of comparative annotation performance against non-collaborative approaches utilizing a single FR engine and a single database only.