The study of homophily in political networks is often clouded by problems of causal inference. Is it driven by social influence, selection, or some combination of the two? Part of the difficulty in distinguishing mechanisms is the emphasis on “core networks.” Politics tends to be talked about most frequently by those with whom we interact most frequently (Song, 2015), who tend to be similar to us – or, “de facto selectivity” (Mutz, 2006). In this paper we take an alternative tack: what factors influence political discussant selection in a “zero acquaintance” (strangers interacting, such as at a party) setting? Study 1 asked undergraduate respondents to provide perceptions of the partisanship, closed-mindedness, friendliness, and religiosity of a random subset of 50 different photographs of strangers. Respondents were able and willing to apply political stereotypes (e.g., Blacks and younger folks are Democrats, closed-minded and religious people are Republicans) in relatively consistent ways to strangers. Study 2 replicates these findings with a diverse adult sample and a smaller subset of images largely drawn from Study 1. Moreover, Study 2 participants were told that these photographs represented other study participants, and that they would soon engage in an online discussion with one of these strangers. Respondents were then asked to indicate, for each photograph, whether or not they would be willing to have a political conversation with the person in that photograph. These data, in conjunction with respondent self-reports, suggest homophily of various sorts – including gender, racial and partisan – drives political discussant selection.