Background Sleep quality may be an important, yet relatively neglected, predictor of treatment outcome in cognitive‐behavioral therapy (CBT) for anxiety disorders. Specifically, poor sleep quality may impair memory consolidation of in‐session extinction learning. We therefore examined sleep quality as a predictor of treatment outcome in CBT for social anxiety disorder and the impact of d‐cycloserine (DCS) on this relationship. Methods One hundred sixty‐nine participants with a primary diagnosis of DSM‐IV generalized social anxiety disorder were recruited across three sites. Participants were enrolled in 12 weeks of group CBT. Participants randomly received 50 mg of DCS (n = 87) or pill placebo (n = 82) 1 hr prior to sessions 3–7. Participants completed a baseline measure of self‐reported sleep quality and daily diaries recording subjective feelings of being rested upon wakening. Outcome measures including social anxiety symptoms and global severity scores were assessed at each session. Results Poorer baseline sleep quality was associated with slower improvement and higher posttreatment social anxiety symptom and severity scores. Moreover, patients who felt more “rested” after sleeping the night following a treatment session had lower levels of symptoms and global severity at the next session, controlling for their symptoms and severity scores the previous session. Neither of these effects were moderated by DCS condition. Conclusions Our findings suggest that poor sleep quality diminishes the effects of CBT for social anxiety disorder and this relation is not attenuated by DCS administration. Therapeutic attention to sleep quality prior to initiation of CBT and during the acute treatment phase may be clinically indicated.