It remains unclear whether social anxiety interferes with the generation of closeness during initial encounters. We addressed the question of whether perceived closeness between strangers differs as a function of dyad characteristics (i.e., self and partner levels of social anxiety) and social context. We conducted an experiment with 90 participants randomly assigned to either a 45-minute personal disclosure or small-talk dyadic conversation. Multilevel modeling results yielded a 3-way interaction, such that the effect of social anxiety on closeness generated during the interaction was moderated by social anxiety reported by interaction partners and social context. In the personal disclosure condition, perceived closeness was greatest when the most socially anxious individuals interacted with each other. In the small-talk condition, perceived closeness was greatest when the least socially anxious individuals interacted with each other. Across conditions, partners with substantial differences in social anxiety (i.e., mixed dyads) reported relatively less closeness than partners with similar levels of social anxiety. Social anxiety effects were not attributable to depressive symptoms or physical attraction to partners. These findings suggest that neglecting specific qualities of interaction partners and social situational factors may lead to spurious conclusions in understanding interpersonal outcomes related to social anxiety.