An index of the activation of Class A G-protein-coupled receptors (GPCRs) has been trained using interhelix distances from a series of microsecond molecular-dynamics simulations and tested for 268 published X-ray structures. In a three-class model that includes intermediate structures, 63% of the active structures are classified in agreement with the experimental assignment, 81% of the intermediate structures, and 89% of the inactives. An alternative two-state model classifies 94% of the actives and 99% of the inactives correctly. The intermediate structures are distributed 2:1 between actives and inactives. X-ray structures with protein nanobodies give good agreement between the assigned activation state and the predictions of the model, whereby many active nanobody structures are predicted to be weakly active. The five interhelix Cα-Cα distances that occur in the model relate clearly to the established activation mechanism. The model is available as a Python script or via an interactive web page. It can thus be used to classify both experimental and computational GPCR structures.