Despite preliminary evidence that individuals with higher social anxiety tend to avoid eye contact during at least some social encounters, the function of this behavior remains unknown. Cognitive theories of social anxiety suggest that gaze avoidance may function as an attempt to avoid signs of social threat. However, it is unclear whether this behavior is effective for reducing anxiety and, if it is effective, whether this benefit is only present in the short-term. The goal of the current study was to test whether gaze avoidance is effective in reducing anxiety for individuals with higher social anxiety during short social conversations among peers. Participants completed a short social interaction with another undergraduate participant in which eye contact was manipulated halfway through the interaction. Recordings of the interactions were later coded for amount of eye contact; this allowed us to obtain an objective measure of adherence to the manipulation instructions. Participants were instructed to make either more or less eye contact, or continue as before. Being asked to make less eye contact was the most anxiety-provoking condition for participants with higher social anxiety. We propose, in line with previous research on safety behaviors, that avoiding eye contact in an effort to regulate state anxiety is an ineffective strategy over time for individuals with higher social anxiety. Therefore gaze avoidance may be a particularly important safety behavior to target in treatment for social anxiety disorder.