This study examined the process by which a web-based sexual violence (SV) prevention program (i.e., "RealConsent") prevents SV perpetration and increases bystander behaviors. Data from 743 college men who participated in a randomized controlled trial were analyzed. Simple and multiple-mediation models were estimated, using several theoretical constructs to assess the mechanisms through which "RealConsent" produced significant effects on SV perpetration and prosocial bystander or intervening behaviors. The results indicated that knowledge of effective consent for sex, hostility toward women, date rape attitudes, and hyper-gender male ideology significantly mediated the effects of "RealConsent" on SV perpetration in the multiple-mediator model. Furthermore, intentions to intervene significantly mediated the effects of "RealConsent" on prosocial bystander behaviors in the multiple-mediator model. The results show that the "RealConsent" program works to prevent SV perpetration and prosocial bystander behaviors via several theoretically proposed mediators central to the development and content of the program. The results also provide strong evidence that SV and bystander education for college men may benefit from including an explicit focus on decreasing negative norms related to women (e.g., hostility, date rape attitudes, hyper-gender ideology) and through increasing college men's knowledge of consent and intentions to intervene.