Cognitive-behavioral therapy (CBT) is effective for pediatric obsessive-compulsive disorder (OCD), but non-response is common. Brain glutamate (Glu) signaling may contribute to OCD pathophysiology and moderate CBT outcomes. We assessed whether Glu measured with magnetic resonance spectroscopy (MRS) was associated with OCD and/or CBT response. Youths aged 7-17 years with DSM-IV OCD and typically developing controls underwent 3 T proton echo-planar spectroscopic imaging (PEPSI) MRS scans of pregenual anterior cingulate cortex (pACC) and ventral posterior cingulate cortex (vPCC)-regions possibly affected by OCD-at baseline. Controls returned for re-scan after 8 weeks. OCD youth-in a randomized rater-blinded trial-were re-scanned after 12-14 weeks of CBT or after 8 weeks of minimal-contact waitlist; waitlist participants underwent a third scan after crossover to 12-14 weeks of CBT. Forty-nine children with OCD (mean age 12.2± 2.9 years) and 29 controls (13.2±2.2 years) provided at least one MRS scan. At baseline, Glu did not differ significantly between OCD and controls in pACC or vPCC. Within controls, Glu was stable from scan-to-scan. Within OCD subjects, a treatment-by-scan interaction (p=0.034) was observed, driven by pACC Glu dropping 19.5% from scan-to-scan for patients randomized to CBT, with minor increases (3.8%) for waitlist participants. The combined OCD participants (CBT-only plus waitlist-CBT) also showed a 16.2% (p= 0.004) post-CBT decrease in pACC Glu. In the combined OCD group, within vPCC, lower pre-CBT Glu predicted greater post-CBT improvement in symptoms (CY-BOCS; r= 0.81, p= 0.00025). Glu may be involved in the pathophysiology of OCD and may moderate response to CBT.