We present a procedure that (i) automates the homology modeling of mammalian olfactory receptors (ORs) based on the first six three-dimensional (3D) structures of G protein- coupled receptors (GPCRs) available and (ii) performs the docking of odorants on these models, using the concept of colony energy to score the complexes. ORs exhibit low-sequence similarities with other GPCRs and current alignment methods often fail to provide a reliable alignment. Here, we use a fold recognition technique to obtain a robust initial alignment. The procedure is then applied to a human OR we had previously functionally characterized. The results, supported by receptor mutagenesis and functional assays, suggest that antagonists dock in the upper part of the binding pocket whereas agonists dock in the narrow lower part. We propose that the potency of agonists in activating receptors depends on their ability to establish tight interactions with the floor of the binding pocket. We developed a web site that allows the user to upload a GPCR sequence, choose a ligand in a library and obtain the 3D structure of the free receptor and ligand receptor complex (http://genome.jouy.inra.fr/GPCRautomodel).