This study aimed at investigating whether stereotyping could be moderated by situations of accountability. In the present study all participants were solicited to form an impression about a target (a gay man) but experimental participants were also asked to communicate it to an audience: psychology students (known audience: KA) vs. other persons (unknown audience: UkA) whereas the audience was not mentioned in the control condition (nA). Results showed that the level of stereotyping was minimal in the UkA condition, intermediate in the KA condition and highest in the no audience condition (nA). Such a difference in the level of stereotyping corresponded to a different path of on-line information processing about the target. Participants in the UkA condition integrated all the available information in order to avoid stereotypical responses, participants in the KA condition endorsed the audience's stereotype in order to deal with the information and finally participants in the nA condition engaged in a category-based information processing.