Media depictions of individuals using tobacco can affect people’s norm perceptions, stigmatizing beliefs, and, ultimately, their behavior. However, media depictions might vary in their effects. Understanding what factors explain this variation is important for advancing communication theory and promoting positive behavior change. Study 1 paired data from a large-scale content analysis of tobacco-related media content published between 2014-2017 with a rolling cross-sectional survey of young people conducted simultaneously. Findings demonstrated that, while media depictions of individual tobacco use did sometimes affect norm perceptions, variation exists in the significance and direction of these effects. Study 2 sought to test whether behavioral similarity and group identity (people’s perceived and desired similarity to a norm referent), might have contributed to this heterogeneity. Study 2a, a content analysis, identified and coded a corpus of YouTube videos depicting individual tobacco use that could be used to explore these research questions. In studies 2b and 2c, young non-tobacco users rated the similarity of different tobacco use behaviors to cigarette smoking (2b) and their perceived and desired similarity to the referents featured using tobacco in videos from study 2a (2c). Results showed that there was variation in participants’ behavioral similarity ratings and in the referent similarity ratings assigned to different videos, but this variation was limited, and these two variables couldn’t be manipulated separately. In study 2d, young non-tobacco users were randomly assigned to a control condition or one of two video conditions in which they were shown either higher or lower referent similarity videos. Participants’ smoking-related norm perceptions and stigma were then measured. Ultimately, no differences were found between participants in the two video conditions on any of these outcomes. Thus, the hypothesized effects of the referent similarity manipulation were not observed. Some differences were observed between the lower referent similarity and control conditions; however, these differences could not be causally attributed to referent similarity. While study 2d’s hypotheses were largely unsupported, results from the set of studies point to a need for more research into how behavioral similarity and referent similarity shape the effects of individual tobacco use depictions on norm perceptions and stigma.