Command hallucinations represent a special problem for the clinical management of psychosis. While compliance with both non-harmful and harmful commands can be problematic, sometimes in the extreme, active efforts to resist commands may also contribute to their malignancy. Previous research suggests Cognitive Behaviour Therapy (CBT) to be a useful treatment for reducing compliance with harmful command hallucinations. The purpose of this trial was to evaluate whether CBT augmented with acceptance-based strategies from Acceptance and Commitment Therapy could more broadly reduce the negative impact of command hallucinations. Forty-three participants with problematic command hallucinations were randomized to receive 15 sessions of the intervention "TORCH" (Treatment of Resistant Command Hallucinations) or the control, Befriending, then followed up for 6 months. A sub-sample of 17 participants was randomized to a waitlist control before being allocated to TORCH or Befriending. Participants engaged equally well with both treatments. Despite TORCH participants subjectively reporting greater improvement in command hallucinations compared to Befriending participants, the study found no significant group differences in primary and secondary outcome measures based on blinded assessment data. Within-group analyses and comparisons between the combined treatments and waitlist suggested, however, that both treatments were beneficial with a differential pattern of outcomes observed across the two conditions.