The study investigated current police practices employed to identify those with a mental illness in police custody, and to evaluate the predictive utility of the Brief Jail Mental Health Screen (BJMHS) and the Jail Screening Assessment Tool (JSAT). One hundred and fifty detainees were recruited from two police stations in Melbourne, Australia. Measures included the Structured Clinical Interview for DSM-IV-TR, BJMHS and JSAT. Axis-I disorders were compared with police decisions regarding identification of mental illness based on their usual practices. Participants were classified as requiring referral for further mental health evaluation according to the screening tools. Results indicated that current police practices produced high false negatives, with many of those experiencing mental illness not identified. There was no significant difference in performance between BJMHS (AUC 0.722) and JSAT (AUC 0.779) in identifying those with a serious mental illness (p 0.109). However, JSAT performed significantly better at identifying any Axis-I disorder, excluding substance use disorders, as compared with BJMHS (AUC 0.815, vs AUC 0.729; p 0.018). Given the high prevalence of mental illness among detainees, there is a pressing need to introduce standardised screening tools for mental illness in police custody. This can assist the police in managing detainees appropriately and securing mental health services as required.