Abstract Background WHO guidelines recommend the lateral flow urine lipoarabinomannan assay (LF-LAM) for TB diagnosis in hospitalised HIV-positive individuals. The role of LF-LAM among ambulant patients remains less well defined. We investigated the sensitivity of LF-LAM among ambulant HIV-positive adults in primary health clinics in South Africa. Methods We enrolled adults (aged ≥18 y) with CD4 counts of ≤150 cells/mm3 who had not received TB treatment or antiretroviral therapy in the preceding 3 or 6 mo, respectively. Research nurses performed the LF-LAM test on freshly voided urine. Results were compared with a reference standard of positive mycobacterial culture (sputum or urine). Results Of 1505 (54.5% female; median age 37 y; median CD4 count 73 cells/mm3) participants, 973 (64.7%) had a mycobacterial culture result; 105/973 (10.8%) were positive for Mycobacterium tuberculosis. LF-LAM sensitivity was 41.9% (95% CI 32.3 to 51.9%) and 19.0% (95% CI 12.0 to 27.9%) using grade 1+ and grade 2+ cut-off points, respectively. Sensitivity increased with severe immunosuppression and in the presence of poor prognostic indicators (low haemoglobin, body mass index). Conclusions When used as the only TB diagnostic test, LF-LAM sensitivity is suboptimal, particularly using the grade 2+ cut-off. More sensitive tests for TB are needed that can be used in primary care settings.