Frequent testing of individuals at high risk of HIV is central to current prevention strategies. We aimed to determine if HIV self-testing would increase frequency of testing in high-risk gay and bisexual men, with a particular focus on men who delayed testing or had never been tested before.In this randomised trial, HIV-negative high-risk gay and bisexual men who reported condomless anal intercourse or more than five male sexual partners in the past 3 months were recruited at three clinical and two community-based sites in Australia. Enrolled participants were randomly assigned (1:1) to the intervention (free HIV self-testing plus facility-based testing) or standard care (facility-based testing only). Participants completed a brief online questionnaire every 3 months, which collected the number of self-tests used and the number and location of facility-based tests, and HIV testing was subsequently sourced from clinical records. The primary outcome of number of HIV tests over 12 months was assessed overall and in two strata: recent (last test ≤2 years ago) and non-recent (>2 years ago or never tested) testers. A statistician who was masked to group allocation analysed the data; analyses included all participants who completed at least one follow-up questionnaire. After the 12 month follow-up, men in the standard care group were offered free self-testing kits for a year. This trial is registered with the Australian New Zealand Clinical Trials Registry, number ACTRN12613001236785.Between Dec 1, 2013, and Feb 5, 2015, 182 men were randomly assigned to self-testing, and 180 to standard care. The analysis population included 178 (98%) men in the self-testing group (174 person-years) and 165 (92%) in the standard care group (162 person-years). Overall, men in the self-testing group had 701 HIV tests (410 self-tests; mean 4·0 tests per year), and men in the standard care group had 313 HIV tests (mean 1·9 tests per year); rate ratio (RR) 2·08 (95% CI 1·82-2·38; p<0·0001). Among recent testers, men in the self-testing group had 627 tests (356 self-tests; mean 4·2 per year), and men in the standard care group had 297 tests (mean 2·1 per year); RR 1·99 (1·73-2·29; p<0·0001). Among non-recent testers, men in the self-testing group had 74 tests (54 self-tests; mean 2·8 per year), and men in the standard care group had 16 tests (mean 0·7 per year); RR 3·95 (2·30-6·78; p<0·0001). The mean number of facility-based HIV tests per year was similar in the self-testing and standard care groups (mean 1·7 vs 1·9 per year, respectively; RR 0·86, 0·74-1·01; p=0·074). No serious adverse events were reported during follow-up.HIV self-testing resulted in a two times increase in frequency of testing in gay and bisexual men at high risk of infection, and a nearly four times increase in non-recent testers, compared with standard care, without reducing the frequency of facility-based HIV testing. HIV self-testing should be made more widely available to help increase testing and earlier diagnosis.The National Health and Medical Research Council, Australia.