Humans are often shown to cooperate with one another. Most of the mechanisms that foster cooperation among humans rely on reputation, which itself relies on the acquisition of information about other people’s behaviors. Gossip has been proposed as a cheap yet efficient tool to acquire information, and it has largely been proved to be an effective means to foster and maintain cooperation. However, empirical studies supporting this claim have ignored two aspects: (1) they often compared gossip to treatments in which no reputation was available, impeding a direct assessment of whether it is gossip that promotes cooperation or rather the introduction of a reputation system; and (2) they focused on pro‐social gossip (e.g., gossip aimed at helping the receiver), neglecting the impact of other types of gossip. We show here that, in contrast with the widespread notion that gossip promotes cooperation, gossip mostly depletes cooperation compared to first‐hand information. If lying is fruitful for individuals or if a group’s behavior is largely uncooperative, gossip leads to negative reputational information and decreased cooperation.