Background: Paranoia is one of the commonest symptoms of psychosis but has rarely been studied in a population at risk of developing psychosis. Based on existing theoretical models, including the proposed distinction between ‘poor me’ and ‘bad me’ paranoia, we test specific predictions about associations between negative cognition, metacognitive beliefs and negative emotions and paranoid ideation and the belief that persecution is deserved (deservedness). Methods: We used data from 117 participants from the EDIE-2 trial of cognitive behaviour therapy for people at high risk of developing psychosis, comparing them with samples of psychiatric inpatients and healthy students from a previous study. Multi-level modelling was utilised to examine predictors of both paranoia and deservedness, with post-hoc planned comparisons conducted to test whether person-level predictor variables were associated differentially with paranoia or with deservedness. Results: Our sample of ARMS participants was not as paranoid, but reported higher levels of “bad-me” deservedness, compared to psychiatric inpatients. We found several predictors of paranoia and deservedness. Negative beliefs about self were related to deservedness but not paranoia, whereas negative beliefs about others were positively related to paranoia but negatively with deservedness. Both depression and negative metacognitive beliefs about paranoid thinking were specifically related to paranoia but not deservedness. Conclusions: This study provides evidence for the role of negative cognition, metacognition and negative affect in the development of paranoid beliefs, which has implications for psychological interventions and our understanding of psychosis.