People tend to generalize the actions of known group members to new ones when they are both members of the same group. This study was conducted to investigate how the prevalence of specific actions among multiple individuals determines action generalization within social groups. We propose that people rely on the belief that group members work toward a shared goal (i.e., shared-goal belief) to guide action generalization. Consequently, the extent of action generalization may not consistently increase with the sampled prevalence of group members performing the same goal-directed action, resulting in a deviation from graded action generalization (i.e., nongraded action generalization). Experiment 1 revealed that the more participants believed that group members pursued a shared goal, the greater the likelihood that nongraded action generalization would occur. In Experiment 2, experimental manipulation weakened the strength of the shared-goal belief and led to a graded pattern of action generalization with accumulated evidence of action prevalence. These findings suggest that a shared-goal belief within groups significantly shapes action generalization beyond the mere influence of sampled action prevalence. Social groups not only provide a framework for selecting evidence for action generalization but also shape prior beliefs that influence our expectations of others' actions.