Objective To evaluate whether therapeutic mechanisms assumed to explain the effect of cognitive behavioural therapy (CBT) concerning temporomandibular symptoms are confirmed by structural equation modelling. Method Patients were randomly assigned to either an experimental group receiving CBT (N = 41) or a standard therapy control group (N = 31). Subjects were assessed before and after intervention using mediator variables hypothesized according to the CBT model, as well as outcome variables and other socio‐demographic and clinical measures. Results The results confirm that the effect of treatment on pain intensity was partially mediated by distress, catastrophizing, perceived control, distraction, and mental self‐control. The self‐medication frequency was partially mediated by distraction. Pain interference was partially mediated by distress, distraction, and mental self‐control. Reduction in the number of painful points on palpation was partially explained by distress, although in this case, there was a significant direct effect of treatment not mediated by other variables. Conclusions The results could set the principles for the development of more efficient and effective cognitive behavioural interventions for chronic pain. Key Practitioner Message Cognitive‐behavioural models fundamentals for chronic pain interventions (and concretely in TMD patients) are tested and partially confirmed, providing experts and professionals with specific knowledge to be transferred to clinical practice. The article provide key messages for the psychologists (and other professionals) working on pain management, identifying those strategies directly linked with the improvement of outcome measures. The results provide evidence‐based orientations for the future application of the CBT to chronic pain, thereby making interventions more efficient as they would only be focused on the most effective technique.