Cannabinoids in current use such as nabilone activate both CB1 and CB2 receptors. Selective CB2 activation may provide some of the therapeutic effects of cannabinoids, such as their immuno-modulatory properties, without the psychoactive effects of CB1 activation. Therefore, cannabinoid CB2 receptors represent an attractive target for drug development. However, selective and potent CB2 agonists remain in development. CB1 and CB2 differ considerably in their amino acid sequence and tertiary structures. Therefore, clinical development of potent and selective CB2 agonists is probable. Mutational and ligand binding studies, functional mapping, and computer modelling have revealed key residues and domains in cannabinoid receptors that are involved in agonist and antagonist binding to CB1 and CB2. In addition, CB2 has undergone more rapid evolution, and results for ligand binding and efficacy cannot be automatically extrapolated from rat or mouse CB2 to human. Furthermore, loss of CB1 affinity is a crucial property for CB2-selective ligands, and although rat CB1 is 97% homologous with human CB1, critical differences do exist, with potential for further exploitation in drug design. In this paper we briefly review previous cannabinoid receptor models and mutation/binding studies. We also review binding affinity ratios with respect to CB1 and CB2. We then employ our own models to illustrate key cannabinoid receptor residues and binding subdomains that are involved in these differences in binding affinities and discuss how these might be exploited in the development of CB2 specific ligands. Published reports for species specific binding affinities for CB2 are scarce, and we argue that this needs to be corrected prior to the progression of CB2 agonists from pre-clinical to clinical research.