Comparative studies using lysophosphatidic acid (LPA) and the synthetic agonist, oleoyl-methoxy glycerophosphothionate (OMPT), in cells expressing the LPA3 receptor revealed differences in the action of these agents. The possibility that more than one recognition cavity might exist for these ligands in the LPA3 receptor was considered. We performed agonist docking studies exploring the whole protein to obtain tridimensional details of the ligand-receptor interaction. Functional in cellulo experiments using mutants were also executed. Our work includes blind docking using the unrefined and refined proteins subjected to hot spot predictions. Distinct ligand protonation (charge -1 and -2) states were evaluated. One LPA recognition cavity is located near the lower surface of the receptor close to the cytoplasm (Lower Cavity). OMPT displayed an affinity for an additional identification cavity detected in the transmembrane and extracellular regions (Upper Cavity). Docking targeted to Trp102 favored binding of both ligands in the transmembrane domain near the extracellular areas (Upper Cavity), but the associating amino acids were not identical due to close sub-cavities. A receptor model was generated using AlphaFold3, which properly identified the transmembrane regions of the sequence and co-modeled the lipid environment accordingly. These two models independently generated (with and without the membrane) and adopted essentially the same conformation, validating the data obtained. A DeepSite analysis of the model predicted two main binding pockets, providing additional confidence in the predicted ligand-binding regions and support for the relevance of the docking-based interaction models. In addition, mutagenesis was performed of the amino acids of the two detected cavities. In the in cellulo studies, LPA action was much less affected by the distinct mutations than that of OMPT (which was almost abolished). Therefore, docking and functional data indicate the presence of distinct agonist binding cavities in the LPA3 receptor.