Computational methodologies are used to increase the efficiency of drug discovery process by rendering the design of new drug candidates more rapid and cost-efficient. In silico techniques can be divided in two main groups. Structure-based drug design procedures can be applied (such as docking simulations) if the target is known from experimental (i.e., X-ray crystallographic studies, NMR studies) or theoretical sources (receptor structure built by homology modeling techniques). Otherwise, ligand-based drug design methods can be used (e.g., QSAR or 3D QSAR models, pharmacophoric models) based on the analysis of a number of ligands known to act with a common mechanism of action. Adenosine receptors (ARs) are a family of G-protein coupled receptors (GPCRs) of great interest as targets for therapeutic intervention. Due partly to the lack of reliable adenosine receptor structures, ligand-based drug discovery methods remain the major computational molecular modeling approach applied in the research of new AR ligands. The scope of this review is to summarize the results on pharmacophoric models and 3D QSAR studies concerning AR ligands. In particular, the review will focus on the use of such ligand-based computational techniques for the identification of new AR ligands and/or for their optimization.