The rationale for utilizing the prompt gamma (PG) signal for in vivo proton beam range verification is such that the PG fall-off distribution along the beam path is associated with the dose profile in the Bragg peak (BP) distal fall-off region. Quantitative characterization of this association, particularly with respect to the BP, is of great importance to assess its limitation and aid in the development of a clinically reliable PG imaging system to maximize PG detection. In this work we investigate the angular dependence of PG detection with respect to the BP for in vivo beam range verification in proton radiation therapy. Geant4 Monte Carlo simulations have been used to study the energy spectral and spatial characteristics of the PG signal from high-energy proton beam irradiations. A cylindrical water phantom (φ30 cm × 50 cm) with an ideal detecting cylinder (φ100 cm × 50 cm) coaxially surrounding the phantom has been used in the simulation. The angular dependence of PG detection as a function of beam energy and PG energy has been characterized with respect to the BP. Our results show that there exists an angular preference for PG detection, which has a strong dependence on the beam energy. As the beam energy increases, the longitudinal angular preference for PG detection becomes increasingly backward with respect to the BP position. This implies that the detector with sufficient longitudinal angular coverage is desired for the BP tracking, especially for the Spread-Out Bragg Peak tracking.