OBJECTIVES: An intervention developed through participatory crowdsourcing methods increased HIV self-testing among men who have sex with men [MSM; relative risk (RR) = 1.89]. We estimated the long-term impact of this intervention on HIV transmission among MSM in four cities (Guangzhou, Shenzhen, Jinan and Qingdao). METHODS: A mathematical model of HIV transmission, testing and treatment among MSM in China was parameterized using city-level demographic and sexual behaviour data and calibrated to HIV prevalence, diagnosis and antiretroviral therapy (ART) coverage data. The model was used to project the HIV infections averted over 20 years (2016-2036) from the intervention to increase self-testing, compared with current testing rates. RESULTS: Running the intervention once would avert < 2.2% infections over 20 years. Repeating the intervention (RR = 1.89) annually would avert 6.4-10.7% of new infections, while further increases in the self-testing rate (hypothetical RR = 3) would avert 11.7-20.7% of new infections. CONCLUSIONS: Repeated annual interventions would give a three- to seven-fold increase in long-term impact compared with a one-off intervention. Other interventions will be needed to more effectively reduce the HIV burden in this population.