To predict the response of mostly elderly patients to brief intervention against long-term benzodiazepine (BZD) use delivered in general medical practice from variables measured at baseline in a randomised controlled trial. Logistic regression was used to identify predictors of a complete cessation of BZD intake or a ‘clinically significant reduction’ by a half or more from baseline to 6 months follow-up among 183 patients who received a brief intervention. Candidate predictor variables were: (i) stage of change (ii) level of BZD dependence (iii) whether BZDs were prescribed by the patient's usual general practitioner (GP) or by another medical practitioner; (iv) baseline BZD dosage; (v) type of BZD and (vi) gender. Both cessation and reduction were predicted by who prescribed BZDs, with patients whose medication was prescribed by their usual GP more likely to show a positive response to brief intervention than those whose medication was prescribed by another medical practitioner. Stage of change was a significant predictor of a reduction in BZD use, with patients in the Contemplation stage nearly three times more likely, and those in the Action stage over eight times more likely, to achieve a clinically significant reduction than those in the Precontemplation stage. Patients receiving prescriptions from their usual GP are more likely to cease or reduce BZD intake than those receiving prescriptions from another medical practitioner. In managing patients with long-term use of BZDs, general medical practitioners should consider recording the patient's stage of change and tailoring their intervention on that basis.