Recently, we demonstrated that the peak-end memory bias, which is well established in the context of pain, can also be observed in anxiety: Retrospective evaluations of a frightening experience are worse when peak anxiety is experienced at the end of an episode. Here, we set out to conceptually replicate and extend this finding with rigorous experimental control in a threat of shock paradigm. We induced two intensity levels of anxiety by presenting visual cues that indicated different strengths of electric stimuli. Each of the 59 participants went through one of two conditions that only differed in the order of moderate and high threat phases. As a manipulation check, orbicularis-EMG to auditory startle probes, electrodermal activity, and state anxiety confirmed the effects of the specific threat exposure. Critically, after some time had passed, participants for whom exposure had ended with high threat reported more anxiety for the entire episode than those for whom it ended with moderate threat. Moreover, they ranked their experience as more aversive when compared to other unpleasant everyday experiences. This study overcomes several previous limitations and speaks to the generalizability of the peak-end bias. Most notably, the findings bear implications for exposure therapy in clinical anxiety. •This research convincingly consolidates evidence for the peak-end bias in anxiety.•The ending of a frightening episode determines how it is evaluated retrospectively.•Inducing anxiety with threat of shock provided us with rigorous experimental control.•Physiological and self-report indices corroborate the graded induction of anxiety.•The findings bear relevant implications for exposure therapy in clinical anxiety.