To identify factors which may help or hinder decision-making ability in people with psychosis, we did a systematic review and meta-analysis of their performance on the Iowa and Cambridge Gambling Tasks. Analysis of 47 samples found they had moderately poorer performance than healthy individuals (N = 4264, g = −0.57, 95% confidence interval (CI) −0.66 to −0.48). Few studies (k = 8) used non-psychotic clinical comparator groups, although very low-quality evidence (k = 3) found people with bipolar disorder may perform better. Negative symptoms (k = 13, N = 648, r = −0.17, 95% CI −0.26 to −0.07) and lower IQ (k = 11, N = 525, r = 0.20, 95% CI 0.29–0.10), but not positive symptoms (k = 10, N = 512, r = −0.01, 95% CI −0.11 to 0.08), each had small-moderate associations with poorer decision-making. Lower quality evidence suggested general symptoms, working memory, social functioning, awareness of emotional responses to information, and attentional bias towards gain are associated with decision-making, but not education, executive functioning or overall symptoms. Meta-regression suggested an inverse association between decision-making and depression severity (k = 6, Q = 6.41, R2 100%, p = 0.01). Those taking first-generation (k = 6, N = 305, g = −0.17, 95% CI −0.40 to 0.06, p = 0.147) or low-dose antipsychotics (k = 5, N = 442, g = −0.19, 95% CI −0.44 to 0.06, p = 0.139) had unimpaired decision-making. Although meta-regression found no linear association between dose and performance, non-reporting of the dose was common and associated with larger impairments (k = 46, Q = 4.71, R2 14%, p = 0.03). Those supporting people with psychosis to make decisions, including treatment decisions, should consider the potential effect of these factors. Interventionist-causal trials are required to test whether reducing antipsychotic dose and treating anxiety and depression can improve decision-making in this group.