Cognitive behaviour therapy (CBT) for generalized anxiety disorder produces variable results. It would be valuable to identify which patients respond well to brief CBT and whether increased intensity of therapy improves outcome for those with a poor prognosis. This clinical effectiveness study was designed to address these issues using a prognostic index developed from earlier research. Suitable patients were given five sessions of CBT if they had a good prognosis and either 9 or 15 sessions of CBT if they had a poor prognosis. There was no evidence that increasing the intensity of therapy for poor prognosis patients improved outcome and 60% remained cases at 6 month follow-up. In contrast, only 12% of good prognosis patients remained cases at follow-up despite receiving only brief therapy. Patient characteristics are a more powerful influence on outcome than the length of therapy. Prognostic indices may have an important place in routine clinical practice. Copyright © 2004 John Wiley & Sons, Ltd.