In this paper, we study how communication shapes suspicion. The experiment uses a sender-receiver setup where senders have a low probability of having misaligned incentives with receivers. We focus on the impact of open communication on the receivers' suspicion as measured by the size of the deviation from the senders' recommendation before and after the communication. Overall, communication substantially reduces suspicion, but there are also receivers who become more suspicious during the communication. We disentangle these effects using machine learning methods to analyze the chat logs.