Across six studies, we test whether implicit anti-Black bias shapes White participants’ affiliative responses toward White targets who are comfortable around Blacks as a function of perceived similarity, a process we term implicit homophily. Participants with higher implicit anti-Black bias reported more negative initial affiliative responses toward targets with Black friends than did participants with low implicit bias (Studies 1 – 3). Also, participants with greater implicit anti-Black bias reported less perceived similarity with White targets with Black friends, but not targets believed to be randomly paired with a Black other. Perceived similarity, in turn, statistically mediated the relationship between participants’ implicit bias and affiliative responses to ingroup targets (Study 2). Furthermore, by manipulating participants’ perceived similarity to targets, we found that implicit anti-Black bias predicted their affiliative responses to targets only when they did not already know how similar targets’ outgroup experiences were to their own (Study 3). Providing ecological and behavioral evidence of implicit homophily, women’s implicit anti-Black bias predicted their likelihood of having Facebook friends with Black friends (Study 4). Finally, implicit homophily effects were found to also occur as a function of nonverbal behavioral cues, such that participants’ implicit anti-Black bias predicted affiliation toward targets who were more nonverbally comfortable around Black strangers (Study 5). The subjective comfort implied by targets’ nonverbal behavior was also more important than objective treatment of the Black stranger (Study 6). Across studies, effects persisted above and beyond the effects of implicit pro-White bias (Studies 1, 2, & 4) and explicit racial bias (Studies 1 – 6). Implications for research on stigma by association, extended contact, affiliation, and network formation are discussed.