SETTING: Governmental health facilities performing TB diagnostics in Manicaland, Zimbabwe. OBJECTIVE: To investigate the effect of making Xpert® MTB/RIF the primary TB diagnostic for all patients presenting with presumptive TB on 1) the number of samples investigated for TB, 2) the proportion testing TB-positive, and 3) the proportion of unsuccessful results over time. DESIGN: This retrospective study used data from GeneX-pert downloads, laboratory registers and quality assurance reports between 1 January 2017 and 31 December 2018. RESULTS: The total number of Xpert tests performed in Manicaland increased from 3,967 in the first quarter of 2017 to 7,011 in the last quarter of 2018. Mycobacterium tuberculosis DNA was detected in 4.9-8.6% of the samples investigated using Xpert, with a higher yield in 2017 than in 2018. The overall proportion of unsuccessful Xpert assays due to "no results", errors and invalid results was 6.3%, and highly variable across sites. CONCLUSION: Roll out of more sensitive TB diagnostics does not necessarily result in an increase of microbiologically confirmed TB diagnosis. While the number of samples tested using Xpert increased, the proportion of TB-positive tests decreased. GeneXpert soft- and hardware infrastructure needs to be strengthened to reduce the rate of unsuccessful assays and therefore, costs and staff time.