This meta-analysis examined what revenge feels like. We located 53 manuscripts with 98 independent samples, including 14,238 participants. Revenge was linked to positive affective states (r = .19, [.12, .25]) but not negative affective states (r = .06, [-.06, .17]). Specifically, revenge was pleasant when both enacted and envisioned, but not when observed. Revenge was pleasant across assessment time and measurement type. However, this pleasure subsided with time, suggesting that revenge was sweeter during the act than shortly after. Intergroup revenge was sweeter than interpersonal one, and revenge was most strongly related to gratification and empowerment. Findings were consistent across publication status, preregistration presence, participants’ age, gender, and race, as well as study type and setting. Conversely, revenge was unpleasant when envisioned and observed, but not when enacted. The link between revenge and negative affective states decreased over the last fifty years and was moderated by the preregistration presence, showing that revenge was bitter in non-preregistered studies but not in preregistered ones. Furthermore, revenge was unrelated to anger, anxiety, and sadness but associated with feelings of hostility instead. Theoretically, these results indicate that revenge is particularly pleasant during the act, emphasizing the rewarding character of retaliatory actions and supporting the reinforcement model of revenge. Practically, they suggest that imagery rescripting can be emotionally beneficial to the victims of transgressions.