Research attempting to train individuals to detect deception has demonstrated small, inconsistent effects. The methods with which previous studies have trained participants and the cues to deception used in these training programs may be partially responsible for these findings. The current study investigated the effectiveness of a novel, interactive training program for deception detection which used an interview setting with virtual humans and an intelligent tutoring system. This training program was compared to a non-interactive training program. Participants were trained on eight indirect, impression-based cues that have been empirically demonstrated to be diagnostic of deception. Results demonstrated that indirect, impression-based cues are effective in training for deception detection. Additionally, the interactive training program proved to be just as effective as the training program that was not interactive. Implications for future deception detection training studies will be discussed.