BACKGROUND: Initial patient perceptions of a treatment's credibility, and expectations for improvement, have been shown to predict to a small degree outcomes for common mental health disorders. This study aimed to discover: how patients with psychosis initially perceive psychological therapy's credibility and likely success; whether there are predictors of these first views; and, primarily, if such ideas predict improvement in persecutory delusions. METHODS: We analysed first therapy session data on credibility and expectancy from 195 patients with non-affective psychosis taking part in clinical trials treating persecutory delusions. Baseline assessments before randomisation to therapy were used to identify potential predictors of credibility and expectancy. First session credibility and expectancy scores were tested as predictors of persecutory delusion severity six months later. RESULTS: Most patients were optimistic about therapy's potential. Baseline delusion severity did not predict credibility, β = -0.02, p = .742, or expectancy, β = -0.03, p = .632. Higher psychological well-being predicted higher levels of expectancy, β = 0.23, p = .001. Higher levels of credibility, β = -0.17, p = .021, and expectancy, β = -0.17, p = .020, predicted lower severity of persecutory delusions six months later. Credibility, β = 0.00, p = .970, and expectancy, β = -0.21, p = .074, did not significantly predict uptake of therapy sessions. CONCLUSIONS: Treatment credibility and expectancy may account for a small proportion of improvement when psychological interventions are used to treat severe paranoia. The results are comparable to those for common mental health disorders. There was little prediction of patient variability in credibility and expectancy. It will be helpful to understand a patient's initial views on intervention so that any concerns can be addressed.