BACKGROUND Most studies examining processes of change in psychological therapy for social anxiety disorder (SAD) have analysed data from randomised controlled trials in research settings. METHODS To assess whether these findings are representative of routine clinical practice, we analysed audit data from two samples of patients who received Cognitive Therapy for SAD (total N = 271). Three process variables (self-focused attention, negative social cognitions, and depressed mood) were examined using multilevel structural equation models. RESULTS Significant indirect effects were observed for all three variables in both samples, with negative social cognitions showing the strongest percent mediation effect. 'Reversed' relationships, where social anxiety predicted subsequent process variable scores, were also supported. CONCLUSIONS The findings suggest the processes of change in this treatment may be similar between research trials and routine care.