Background Cannabis use shows a robust dose-dependent relationship with psychosis risk among the general population. Despite this, it has been difficult to link cannabis use with risk for transitioning to a psychotic disorder among individuals at ultra-high risk (UHR) for psychosis. The present study examined UHR transition risk as a function of cannabis use characteristics which vary substantially between individuals including age of first use, cannabis abuse severity and a history of cannabis-induced attenuated psychotic symptoms (APS). Method Participants were 190 UHR individuals (76 males) recruited at entry to treatment between 2000 and 2006. They completed a comprehensive baseline assessment including a survey of cannabis use characteristics during the period of heaviest use. Outcome was transition to a psychotic disorder, with mean time to follow-up of 5.0 years (range 2.4-8.7 years). Results A history of cannabis abuse was reported in 58% of the sample. Of these, 26% reported a history of cannabis-induced APS. These individuals were 4.90 (95% confidence interval 1.93-12.44) times more likely to transition to a psychotic disorder (p = 0.001). Greater severity of cannabis abuse also predicted transition to psychosis (p = 0.036). However, this effect was mediated by higher abuse severity among individuals with a history of cannabis-induced APS. Conclusions Findings suggest that cannabis use poses risk in a subpopulation of UHR individuals who manifest cannabis-induced APS. Whether this reflects underlying genetic vulnerability requires further study. Nevertheless, findings reveal an important early marker of risk with potentially significant prognostic utility for UHR individuals.