The present paper proposes a novel method for 3D-model watermarking for indexing. The proposed approach is based on the use of a Generalized Radon Transformation. More specifically, the Cylindrical Integration Transform (CIT) is initially applied to the 3D models in order to produce descriptor vectors. At the same time a watermarking technique, based on CIT is used in order to embed a specific model identifier in the nodes of the 3D model. This identifier links the model to its descriptor vector, which is extracted only once and stored in a database. Every time this model is employed as a query model, watermark detection is used so as to retrieve the corresponding identifier and further the descriptor vector, which can be further used in a matching algorithm. The proposed techniques are evaluated experimentally in terms of both watermarking efficiency and content-based retrieval performance.