Patients with multi-episode bipolar and psychotic disorders have a high prevalence of substance use disorders, with negative consequences. A brief, easily administered screening test such as the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST) is needed to identify those at risk in order to intervene appropriately. However, the ASSIST has not yet been validated in this population. This article aims to determine the validity and reliability of the ASSIST in detecting substance use disorders in patients with multi-episode bipolar and psychotic disorders. Western Cape Province, South Africa. The Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Health Disorders, 4th Edition (DSM-IV) Axis I Disorders (SCID-I) was used as the gold standard for detecting substance abuse and dependence. Cronbach's alpha was used to determine the internal consistency of the ASSIST, and receiver operating characteristic analysis was used to evaluate its screening properties. Optimal cut off scores were calculated to maximise sensitivity and specificity. A total substance involvement lifetime score of ≥13 was found to have optimal sensitivity and specificity of just over 74%. The optimal cutoff score for alcohol was ≥4 and for cannabis, methamphetamine, and 'other drugs' was ≥3. The area under the curve was 0.7 or above for both the total and specific substance involvement scores. The ASSIST is a psychometrically sound screening test for substance use disorders in patients with multi-episode bipolar and psychotic disorders. This is the first study to validate the ASSIST in this population.