G protein–coupled receptors (GPCRs) play an important role in physiology and disease and represent the most productive drug targets. Orphan GPCRs, with their endogenous ligands unknown, were considered a source of drug targets and consequently attract great interest to identify their endogenous cognate ligands for deorphanization. However, a contrary view to the ubiquitous existence of endogenous ligands for every GPCR is that there might be a significant overlooked fraction of orphan GPCRs that function constitutively in a ligand-independent manner only. Here, we investigated the evolution of the bombesin receptor–ligand family in vertebrates in which one member—bombesin receptor subtype-3 (BRS3)—is a potential orphan GPCR. With analysis of 17 vertebrate BRS3 structures and 10 vertebrate BRS3 functional data, our results demonstrated that nonplacental vertebrate BRS3 still connects to the original ligands—neuromedin B (NMB) and gastrin-releasing peptide (GRP)—because of adaptive evolution, with significantly changed protein structure, especially in three altered key residues (Q127R, P205S, and R294H) originally involved in ligand binding/activation, whereas the placental mammalian BRS3 lost the binding affinity to NMB/GRP and constitutively activates Gs/Gq/G12 signaling in a ligand-independent manner. Moreover, the N terminus of placental mammalian BRS3 underwent positive selection, exhibiting significant structural differences compared to nonplacental vertebrate BRS3, and this domain plays an important role in constitutive activity of placental mammalian BRS3. In conclusion, constitutively active BRS3 is a genuinely orphan GPCR in placental mammals, including human. To our knowledge, this study identified the first example that might represent a new group of genuinely orphan GPCRs that will never be deorphanized by the discovery of a natural ligand and provided new perspectives in addition to the current ligand-driven GPCR deorphanization. The current model on G protein–coupled receptors (GPCRs) is that they have at least one endogenous ligand for the activation of the heterotrimeric G proteins that controls lots of physiological functions. Currently, 140 out of approximately 800 GPCRs are referred to as orphan because of a lack of knowledge about their endogenous ligands, and they attract great interest in major medical institutes and pharmaceutical companies. Can a “ligand-receptor” model apply to all the GPCRs? Our study shows that it probably cannot. By studying the evolution of a classic orphan GPCR–bombesin receptor subtype-3 (BRS3), we show that placental mammalian BRS3 is constitutively active, in contrast to nonplacental vertebrate BRS3, which actually connects to its original ligands—neuromedin B (NMB) and gastrin-releasing peptide (GRP). Protein structure analysis and experiment data also suggest that placental mammalian BRS3, including human BRS3, lost connection with its original ligands during placental mammalian evolution. Therefore, we propose a new point of view that a considerable number of orphan GPCRs do not have endogenous ligands and might represent a new group of GPCRs that are genuinely orphan GPCRs and will never be deorphanized by discovery of a natural ligand. This new finding will provide new perspectives in current ligand-driven GPCR deorphanization.