<sec> <title>SETTING</title> The proportion of bacteriologically confirmed tuberculosis (TB) in China has decreased far below the worldwide average. </sec> <sec> <title>OBJECTIVE</title> To investigate whether stepwise measures to ensure sputum quality can improve the rate of bacteriologically confirmed TB. </sec> <sec> <title>DESIGN</title> We enrolled 980 adults with suspected TB from three counties in China during 2017 for this multicentre randomised controlled trial. Half the participants (n = 490) were randomly assigned to intervention groups that received instructions by a study nurse, and sputum induction, if necessary. In the remaining 490 patients, sputum samples were collected without observation. The primary outcome was the proportion of patients detected as bacteriologically positive on smear, culture or molecular assays (EasyNAT or Xpert). </sec> <sec> <title>RESULTS</title> Bacteriological confirmation rates were significantly higher in the intervention than in the control group: overall (159/490 [32%] vs. 122/490 [25%]; P = 0.009); confirmation using smear (17% vs. 11%; P = 0.010); confirmation using culture (28% vs. 21%; P = 0.021); and confirmation using molecular assays (27% vs. 18%; P = 0.001). Most of the improvement was in patients who received instruction alone, while improvement was greatest in younger patients (adjusted odds ratio 1.27, 95%CI 1.05-1.53 per 10 years). </sec> <sec> <title>CONCLUSIONS</title> If implemented effectively in resource-limited primary care clinics, our simple stepwise procedure combining instruction and sputum induction could increase the proportion of bacteriologically confirmed TB significantly. </sec>.