Content classification performed by end users is spreading through the web. Most of the work done so far is related to the hypermedia web. In spite of that, there is a growing mass of 3D content that is not associated to high-level information. This situation prevents any advanced use of the data available, such as searching, extraction or automatic manipulation of such content. In the last few years there have been a number of approaches for adding metadata to 3D content, but most of them use a top-down, designer driven, classification. This work proposes a different approach based on a bottom-up, user driven, classification. A web based architecture for tagging and browsing 3D objects and worlds is presented and applied to a case study related to the cultural heritage domain.