Introduction. Previous research has determined that adolescents, compared to other age groups, are the most prone to risky online behavior (Ciboci et al., 2020; Feng & Xie, 2014; Shin & Kang, 2016; Steijn et al., 2016). The subject of this research is risky adolescent behavior through which they establish contact with unknown persons via social networks, which according to the online risk categorization by authors Livingstone and Stoilova (2021) can be classified as contact risk. Establishing contact with unknown person via social network is operationalized as engaging in private online communication with unknown, potentially dangerous person via social network and adding unknown person (as friend or follower) to a social network profile. Since current research of risky adolescent online behavior cannot fully explain such behaviors and is focused mainly on cognitive factors (attitudes towards risky behavior, online privacy concerns) that are related to them, the goal of the doctoral research was to examine the Prototype Willingness Model (Gerrard and Gibbons, 1995; Gibbons et al., 2008), which uses non-cognitive factors, for explanation of adolescents' tendency to establish unknown contacts through social networks. Although previous research (Gool et al., 2015; Ouytsel et al., 2020; Paluckaitė & Žardeckaitė-Matulaitienė, 2021; Walrave et al., 2015) had certain methodological issues, they showed that the Prototype Willingness Model can explain some other adolescents' risky online behaviors. In accordance with the Prototype Willingness Model, the examined model consisted of two paths that represent two processes that are supposed to underlie decision-making - a socially reactive and a rational path that together lead to risky behavior. The special contribution of this research is an attempt to improve the methodology of the existing measures of the Prototype Willingness Model constructs and the inclusion of additional variables related to the characteristics of the adolescents (their age and gender, personality traits; extraversion and sensation seeking), their emotional and behavioral problems, and their involvement in risky behaviors in real life. Methodology. The doctoral research used a short-term longitudinal design with measurements nine weeks apart so that risky online behavior could be assessed twice. Self-assessment measures for adolescents were used, and the research was conducted during school classes. In this research existing instruments were used; Balanced Inventory of Desirable Responding (Paulhus, 1991), Short Extraversion Scale (Mlačić & Goldberg, 2007), Brief Sensation Seeking Scale (Hoyle et al., 2002), three subscales (hyperactivity, behavioral problems and emotional difficulties) of the Strengths and Difficulties Questionnaire (SDQ, Goodman, 1997) and the Scale of Risky Adolescents' Behavior (Ricijaš et al., 2010). Scales that were constructed for the purposes of this research were also used; Scale of Sociodemographic Characteristics and Digital Devices Use, Attitudes Towards Risky Online Behavior Scale, Subjective Social Norms Scale, Prototype Scale, Behavioral Willingness Scale, Behavioral Intention Scale and Risky Online Behavior Scale. The sample consisted of 667 adolescents, 14 to 20 years old (M = 16.30, SD = 1.181); students of the first, second, third and fourth grades of secondary schools from city of Osijek, Croatia. Results and discussion. The SPSS 26.0 program with PROCESS and AMOS packages was used in data processing. The results showed that boys have more positive attitudes towards risky online behavior, more positively perceived parental norms and higher prototype favorability than girls. Extraversion was a significant negative predictor of perception of parental and peers norms and prototype similarity, and sensation seeking was a significant positive predictor of attitudes towards risky online behavior, perceived peer norms, prototype favorability and similarity. Emotional and behavioral problems were significant positive predictors of prototype similarity, and risky behavior in real life was a positive predictor of attitudes toward risky online behavior and perceived parental norms. Past risky online behavior was the best positive predictor of attitudes towards risky online behavior, perceived parental and peer norms and prototype favorability and similarity, and it seemed to have a mediating role in the association of emotional and behavioral problems of adolescents with their attitudes towards risky online behavior, perceived parental and peer norms and prototype similarity. Adolescents who had more pronounced positive attitude towards risky online behavior, more pronounced positive perception of social norms and higher prototype favorability and similarity established unknown online contacts more often. Behavioral intention and behavioral willingness were mediators in the relationships of their attitudes and perception of social norms with their risky online behavior, and behavioral willingness was a mediator in the relationships of prototype favorability and similarity with their risky online behavior. Also, adolescents who had a more pronounced behavioral willingness, compared to those adolescents whose behavioral willingness was less pronounced, established unknown online contacts more often, with the mediating effect of behavioral intention in this relationship. The variables included in the rational path explained a larger share of the variance of adolescents' engagement in unknown online contacts (37.0%) than the variables included in the socially reactive path of the model (29.7%). Also, all variables together, the rational and socially reactive path together, explained a larger share of the variance of adolescents' engagement in unknown online contacts (38.5%) than each of them individually. Conclusion. Behavioral intention was best predictor of adolescent risky online behavior. However, in accordance with the Prototype Willingness Model, the results show that establishing risky online contact is not always planned, but also includes a minor unintentional component. Our results indicate that a large number of adolescents establish unknown online contacts, which points to the need for continuous education of experts and parents. Information about the frequency of establishing unknown online contacts, as well as the adolescents' attitudes towards such behavior and other factors that influence its occurrence, give us guidelines when creating the content of educational and prevention programs. There are also some methodological shortcomings of this research that can help with future research planning in this area, and they relate to the used sample, the instruments, the research procedure and potential additional connections with other constructs (adolescents’ depression, life satisfaction, social anxiety, etc.). One of the major shortcomings is the fact that the research was not conducted on a nationally representative sample, so caution is needed when generalizing the findings.