Background Despite growing research, it remains unclear if cannabis use is associated with additive cognitive impairment in people with psychotic illness and whether exposure in early adolescence is associated with poorer cognitive performance in adulthood. Methods This cross-sectional study of a nationally representative sample of 1199 adults with psychotic illness compared current cognition (digit symbol coding) of 297 current users of cannabis (used in the past year), 460 past users (used previously but not in the past year) and 442 non-users (never used). Multiple logistic regression was used to examine whether cognitive performance of cannabis-user groups varied by exposure age and diagnosis (non-affective/affective psychoses). Results Unadjusted analysis showed current cannabis users had significantly higher odds of impaired cognitive function compared to non-users (odds ratio = 1.52, 95%CI = 1.04 − 2.22). After adjusting for potential confounders, differences between the three groups were not significant. Exposure age was not significant in adjusted analysis. In participants with nonaffective psychoses, cognitive ability of current cannabis users did not differ from non-users. However, in participants with affective psychoses, using cannabis in the last year was a significant predictor of impaired cognitive function (odds ratio = 2.25, 95%CI = 1.05 − 4.84). Conclusion Among people with psychotic illness, there was no significant difference in cognitive function between current, past and non-users of cannabis. However, when we compared cognitive performance of the three cannabis groups by diagnostic grouping, current cannabis use had a significant negative relationship with cognitive function in people with affective psychoses, but not in those with non-affective psychoses. This finding requires replication and further investigation.