To determine if the provision of rapid HIV testing to men who have sex with men attending a health service would increase their frequency of HIV testing over time. Design Non-blinded randomised controlled trial. Setting Public sexual health service in Australia. Participants Men who reported having a male sexual partner within the previous year and an HIV test within the previous two years. Of 400 men entered, 370 (92.5 ) completed the study. Interventions Men attending the service between September 2010 and March 2011 were randomised 1:1 to either ongoing access to rapid HIV testing obtained with finger prick or to conventional HIV serology with venepuncture, over 18 months. Main outcome measure The incidence of all HIV testing after enrolment, including testing outside the study clinic, analysed by intention to treat. Results Of 200 men randomised to the rapid testing arm, 196 were followed for 288 person years. Of 200 men randomised to the conventional testing arm, 194 were followed for 278 person years. Median time since the last HIV test was six months for both arms. Men in the rapid test arm had 469 tests (mean 1.63 tests a year), and men in the conventional test arm had 396 tests (mean 1.42 tests a year); incidence rate ratio 1.15, 95 confidence interval 0.96 to 1.38; P=0.12. In a post hoc analysis, rates of initial HIV testing during follow-up were 1.32 and 1.01 tests a year, respectively (1.32, 1.05 to 1.65; P=0.02). Conclusions Provision of access to rapid HIV testing in a health service did not result in a sustained increase over time in HIV testing by men who have sex with men; however, the rate of initial HIV testing did increase by a third. Further research is required to determine how to achieve sustained increases in the frequency of HIV testing by populations at risk.