BACKGROUND: Cognitive therapy for social anxiety disorder (CT-SAD) is recommended by NICE (2013) as a first-line intervention. Take up in routine services is limited by the need for up to 14 ninety-min face-to-face sessions, some of which are out of the office. An internet-based version of the treatment (iCT-SAD) with remote therapist support may achieve similar outcomes with less therapist time. METHODS: 102 patients with social anxiety disorder were randomised to iCT-SAD, CT-SAD, or waitlist (WAIT) control, each for 14 weeks. WAIT patients were randomised to the treatments after wait. Assessments were at pre-treatment/wait, midtreatment/wait, posttreatment/wait, and follow-ups 3 & 12 months after treatment. The pre-registered (ISRCTN 95 458 747) primary outcome was the social anxiety disorder composite, which combines 6 independent assessor and patient self-report scales of social anxiety. Secondary outcomes included disability, general anxiety, depression and a behaviour test. RESULTS: CT-SAD and iCT-SAD were both superior to WAIT on all measures. iCT-SAD did not differ from CT-SAD on the primary outcome at post-treatment or follow-up. Total therapist time in iCT-SAD was 6.45 h. CT-SAD required 15.8 h for the same reduction in social anxiety. Mediation analysis indicated that change in process variables specified in cognitive models accounted for 60% of the improvements associated with either treatment. Unlike the primary outcome, there was a significant but small difference in favour of CT-SAD on the behaviour test. CONCLUSIONS: When compared to conventional face-to-face therapy, iCT-SAD can more than double the amount of symptom change associated with each therapist hour.