The underlying mechanisms of the effectiveness of cognitive behavioural interventions for chronic pain need further clarification. The role of, and associations between, pain-related psychological flexibility (PF) and pain catastrophizing (PC) were examined during a randomized controlled trial on internet-based Acceptance & Commitment Therapy (ACT) for chronic pain. We assessed (1) the unique and combined indirect effects of PF and PC on outcomes, and (2) the causality of relations between PF, PC and the primary outcome pain interference in daily life (MPI) during ACT. A total of 238 pain sufferers were allocated to either ACT, a control condition on Expressive Writing, or a waiting list condition. Non-parametric cross-product of coefficients mediational analyses and cross-lagged panel designs were applied. Compared to control conditions, both baseline to post-intervention changes in PF and PC seemed to uniquely mediate baseline to three-month follow-up changes in pain interference and psychological distress. Only PF mediated changes in pain intensity. Indirect effects were twice as large for PF (κ² = .09–.19) than for PC (κ² PCS = .05–.10). Further assessment of changes during ACT showed, however, that only PF, and not PC, predicted subsequent changes in MPI, while early and late changes in both PF and PC predicted later changes in each other. In conclusion, only PF functioned as a direct, causal working mechanism during ACT, with larger indirect effects that occurred earlier than changes in PC. Additionally, PC may function as an indirect mechanism of change during ACT for chronic pain via its direct influence on PF.