The World Health Organization (WHO) recommends systematic symptom screening for tuberculosis (TB). However, TB prevalence surveys suggest that this strategy does not identify millions of TB patients, globally. Undiagnosed or delayed diagnosis of TB contribute to TB transmission and exacerbate morbidity and mortality. We conducted a cluster-randomized trial of large urban and rural primary healthcare clinics in 3 provinces of South Africa to evaluate whether a novel intervention of targeted universal testing for TB (TUTT) in high-risk groups diagnosed more patients with TB per month compared to current standard of care (SoC) symptom-directed TB testing. Sixty-two clinics were randomized; with initiation of the intervention clinics over 6 months from March 2019. The study was prematurely stopped in March 2020 due to clinics restricting access to patients, and then a week later due to the Coronavirus Disease 2019 (COVID-19) national lockdown; by then, we had accrued a similar number of TB diagnoses to that of the power estimates and permanently stopped the trial. In intervention clinics, attendees living with HIV, those self-reporting a recent close contact with TB, or a prior episode of TB were all offered a sputum test for TB, irrespective of whether they reported symptoms of TB. We analyzed data abstracted from the national public sector laboratory database using Poisson regression models and compared the mean number of TB patients diagnosed per clinic per month between the study arms. Intervention clinics diagnosed 6,777 patients with TB, 20.7 patients with TB per clinic month (95% CI 16.7, 24.8) versus 6,750, 18.8 patients with TB per clinic month (95% CI 15.3, 22.2) in control clinics during study months. A direct comparison, adjusting for province and clinic TB case volume strata, did not show a significant difference in the number of TB cases between the 2 arms, incidence rate ratio (IRR) 1.14 (95% CI 0.94, 1.38, p = 0.46). However, prespecified difference-in-differences analyses showed that while the rate of TB diagnoses in control clinics decreased over time, intervention clinics had a 17% relative increase in TB patients diagnosed per month compared to the prior year, interaction IRR 1.17 (95% CI 1.14, 1.19, p < 0.001). Trial limitations were the premature stop due to COVID-19 lockdowns and the absence of between-arm comparisons of initiation and outcomes of TB treatment in those diagnosed with TB. Our trial suggests that the implementation of TUTT in these 3 groups at extreme risk of TB identified more TB patients than SoC and could assist in reducing undiagnosed TB patients in settings of high TB prevalence. South African National Clinical Trials Registry DOH-27-092021-4901.