Purpose There is considerable variation in epidemiology and clinical course of psychotic disorders across social and geographical contexts. To date, very little data are available from low-and middle-income countries. In sub-Saharan Africa, most people with psychoses remain undetected and untreated, partly due to lack of formal health care services. This study in rural South Africa aimed to investigate if it is possible to identify individuals with recent-onset psychosis in collaboration with traditional health practitioners (THPs). Methods We developed a strategy to engage with THPs. Fifty THPs agreed to collaborate and were asked to refer helpseeking clients with recent-onset psychosis to the study. At referral, the THPs rated probability of psychosis ("maybe disturbed" or "disturbed"). A two-step diagnostic procedure was conducted, including the self-report Community Assessment of Psychic Experiences (CAPE) as screening instrument, and a semi-structured interview using the Schedules for Clinical Assessment in Neuropsychiatry (SCAN). Accuracy of THP referrals, and test characteristics of the THP rating and the CAPE were calculated. Results 149 help-seeking clients were referred by THPs, of which 44 (29.5%) received a SCAN DSM-IV diagnosis of psychotic disorder. The positive predictive value of a THP "disturbed" rating was 53.8%. Test characteristics of the CAPE were poor. Conclusion THPs were open to identifying and referring individuals with possible psychosis. They recognized "being disturbed" as a condition for which collaboration with formal psychiatric services might be beneficial. By contrast, the CAPE performed poorly as a screening instrument. Collaboration with THPs is a promising approach to improve detection of individuals with recent-onset psychosis in rural South Africa.