Xpert MTB/RIF is approved for use in tuberculosis (TB) and rifampicin-resistance diagnosis. However, data are limited on the impact of Xpert under routine conditions in settings with high TB burden. A pragmatic prospective cluster-randomised trial of Xpert for all individuals with presumptive (symptomatic) TB compared to the routine diagnostic algorithm of sputum microscopy and limited use of culture was conducted in a large TB/HIV primary care clinic. The primary outcome was the proportion of bacteriologically confirmed TB cases not initiating TB treatment by 3 mo after presentation. Secondary outcomes included time to TB treatment and mortality. Unblinded randomisation occurred on a weekly basis. Xpert and smear microscopy were performed on site. Analysis was both by intention to treat (ITT) and per protocol. Between 7 September 2010 and 28 October 2011, 1,985 participants were assigned to the Xpert (n = 982) and routine (n = 1,003) diagnostic algorithms (ITT analysis); 882 received Xpert and 1,063 routine (per protocol analysis). 13% (32/257) of individuals with bacteriologically confirmed TB (smear, culture, or Xpert) did not initiate treatment by 3 mo after presentation in the Xpert arm, compared to 25% (41/167) in the routine arm (ITT analysis, risk ratio 0.51, 95% CI 0.33–0.77, p = 0.0052). The yield of bacteriologically confirmed TB cases among patients with presumptive TB was 17% (167/1,003) with routine diagnosis and 26% (257/982) with Xpert diagnosis (ITT analysis, risk ratio 1.57, 95% CI 1.32–1.87, p<0.001). This difference in diagnosis rates resulted in a higher rate of treatment initiation in the Xpert arm: 23% (229/1,003) and 28% (277/982) in the routine and Xpert arms, respectively (ITT analysis, risk ratio 1.24, 95% CI 1.06–1.44, p = 0.013). Time to treatment initiation was improved overall (ITT analysis, hazard ratio 0.76, 95% CI 0.63–0.92, p = 0.005) and among HIV-infected participants (ITT analysis, hazard ratio 0.67, 95% CI 0.53–0.85, p = 0.001). There was no difference in 6-mo mortality with Xpert versus routine diagnosis. Study limitations included incorrect intervention allocation for a high proportion of participants and that the study was conducted in a single clinic.These data suggest that in this routine primary care setting, use of Xpert to diagnose TB increased the number of individuals with bacteriologically confirmed TB who were treated by 3 mo and reduced time to treatment initiation, particularly among HIV-infected participants.