Most existent research on the effects of interpersonal discussions about health campaign messages is based on surveys. In this study, we analysed actual conversations about an HIV/AIDS poster to find out possible effects. Young South African women in 59 dyads (n = 118) participated in conversations about a deliberately puzzling HIV and AIDS poster that cautioned the target group to be faithful to one sexual partner. We measured their comprehension of the poster and beliefs about the message, before and after the conversations. Overall, actual comprehension (AC) was low, and we observed a large discrepancy between actual and perceived comprehension. In general, conversations did not improve AC. It proved to be even more probable that a correct interpretation before a conversation turned into an incorrect interpretation than the other way around. However, having a well-informed conversation partner increased the chance of acquiring adequate subsequent comprehension. We found, in general, that conversations did not decrease undesirable beliefs. One important undesirable belief even became reinforced after the conversations. Conversations among peers might be valuable in health campaigns, but our study shows that intended positive effects do not automatically follow.