The present study examined the impact of peer-like and expert-like agent stereotypes, as operationalized by agent’s image and voice, on learners’ agent perceptions, task-related attitudes, and learning achievement. 56 university freshmen (23 males and 33 females) interacted with either the peer-like agent (female college student) or the expert-like agent (female college lecturer) in computer-based multimedia lesson on C-Programming. Consistent with similarity-attraction hypothesis, expert hypothesis, and interference hypothesis, it was found that: (1) learners assigned higher ratings on lesson enjoyment with peer-like agent than with expert-like agent, (2) female learners assigned higher trust to the lesson presented by expert-like agent that to the lesson presented by peer-like agent, (3) female learners reported less anxiety in learning task with expert-like agent than with peer-like agent, and (4) learners who were more attracted to virtual agent were more likely to score lower in learning achievement. Additionally, results from this study suggest that gender bias affects learner’s perception on virtual agent. Implications are discussed in terms of how stereotypes of expert-like and peer-like agent can be effectively utilized in e-learning environment.