INTRODUCTION: Past research shows that HIV self-testing (HIVST) can increase testing and facilitate more HIV diagnoses relative to clinic testing. However, in the United States, the use of HIVSTs is limited due to concerns that those who use HIVST could be less likely to be linked to care. METHODS: From January 2019 to April 2022, we recruited 811 men who have sex with men (MSM) in the United States who tested infrequently using an online marketing campaign and randomized them 1:1:1 to receive one of the following every 3 months for a year: (1) text message reminders to get tested at a local clinic (control); (2) mailed HIVST kits with access to a free helpline (standard HIVST); and (3) mailed HIVST kits with counselling provided within 24 hours of opening a kit (eTest). Quarterly follow-up surveys assessed HIV testing, sexually transmitted infection (STI) testing, pre-exposure prophylaxis (PrEP) use and sexual risk behaviour. FINDINGS: Eight participants were diagnosed with HIV, and all but one were through HIVST. Participants in either HIVST condition, standard or eTest, had significantly higher odds of any testing (OR = 7.9, 95% CI = 4.9-12.9 and OR = 6.6, 95% CI = 4.2-10.5) and repeat testing (>1 test; OR = 8.5, 95% CI = 5.7-12.6; OR = 8.9, 95% CI = 6.1-13.4) over 12 months relative to the control group. Rates of STI testing and PrEP uptake did not differ across study condition, but those in the eTest condition reported 27% fewer sexual risk events across the study period relative to other groups. CONCLUSIONS: HIVST vastly increased testing, encouraged more regular testing among MSM, and identified nearly all new cases, suggesting that HIVST could diagnose HIV acquisition earlier. Providing timely follow-up counselling after HIVST did not increase rates of STI testing or PrEP use, but some evidence suggested that counselling may have reduced sexual risk behaviour. To encourage more optimal testing, programmes should incorporate HIVST and ship kits directly to recipients at regular intervals.