Introduction and Aims: Insight is a multi-dimensional construct that predicts treatment outcomes of people with mental illness. Research into insight in substance dependent populations is limited and measures of cognitive insight have not been validated for this population. Design and Methods: A cross sectional survey was conducted with residents of nine residential substance dependence treatment facilities in Australia. Cognitive insight was assessed using the Beck Cognitive Insight Scale (BCIS). Psychological distress was assessed using the Kessler 6 (K6). Results: Participants (N = 312) were primarily male (68.6%), with an average age of 37.51 years (SD = 9.85). Methamphetamine (45.2%) and alcohol (35.9%) were the primary substances of use. A confirmatory factor analysis confirmed the two-factor model of the BCIS (CMIN/DF = 2.91, CFI = 0.84). Removing two items from the Self-Reflection subscale improved model fit (CMIN/DF = 2.71, CFI = 0.84, Χ 2[22, n = 312] = 76.43, P < 0.02). Internal consistency analyses indicated acceptable internal reliability (Self-Reflection α = 0.73, Self-Certainty α = 0.72, composite α = 0.75). Self-Certainty scores were significantly higher for participants with a self-reported psychotic disorder (M = 14.95 vs. M = 13.04, P = 0.007). Self-Reflection scores were higher for people experiencing psychological distress (M = 17.57 vs. M = 15.95, P = 0.001). Discussion and Conclusions: We found that a 12-item version of the BCIS had good psychometric properties in this substance-using population. Further research is needed to explore whether insight can predict treatment outcomes for substance use.