As media technology evolves quickly, tailored and targeted communication has emerged as an important strategy in persuasion. One of the most widely used and easiest tactics of tailoring is using an exemplar or persuader character similar to the target audience. However, the effect of character-audience similarity may significantly differ depending on other message features such as how the similar character is shown to behave within the message. This dissertation examines the positive and negative effects of the similarity between the audience and the exemplar character in anti-smoking public service announcements (PSAs), based on demographics and smoking status. Would character-audience similarity still exert positive effect on persuasion even when the target of similarity assessment is shown to behave in socially unacceptable ways? Two secondary data analyses on video anti-smoking PSAs were conducted first. Study 1 examined the main effect of character-audience similarity, finding a significant positive effect of demographic similarity between the smoker character and the audience. Study 2 focused on the secondhand smoking (SHS)-themed PSAs, where the smoker character’s behavior may be seen as irresponsible and immoral by causing serious harm to others. The impact of character-audience similarity depended on the severity of consequences caused by SHS, so that the usually positive effects of character-audience similarity disappeared in PSAs depicting highly severe consequences of SHS. However, the effect was only marginally significant. Informed by these two studies, an experiment (Study 3) was conducted to systematically manipulate and examine the effect of character-audience similarity (Similar vs. Dissimilar), theme of the message (self-harm vs. harm of SHS), and severity of consequences (high vs. low severity). A consistently negative and significant two-way interaction effect between character-audience similarity and theme emerged, suggesting that seeing a similar smoker character harming their own health (self-harm condition) increased engagement with the message and identification with the character, which in turn was associated with greater perceived effectiveness (PE). However, seeing a similar smoker endangering others via secondhand smoking (SHS condition) decreased engagement and identification, causing a boomerang effect on message effectiveness. The results provide valuable guidelines for message design regarding the use of character-audience similarity: When the exemplar character acts in socially undesirable ways, such as endangering others via secondhand smoking, character-audience similarity might backfire, and message designers should avoid tailoring via character-audience similarity. While the studies in this dissertation were limited in the topics of smoking tobacco cigarettes, other behaviors (e.g. drunk driving) may also be subject to similar effects given that driving under the influence may lead to serious negative consequences on innocent others.