BACKGROUND: Female sex workers remain disproportionately affected by HIV. The aim of this study was to determine the effect of risk-differentiated, peer-led support for female sex workers in Zimbabwe on the risk of HIV acquisition and HIV transmission from sex among female sex workers. METHODS: In this cluster randomised, open-label, controlled study, 22 clinics dedicated to female sex workers co-located in government health facilities throughout Zimbabwe were allocated (1:1, through restricted randomisation) to usual care or AMETHIST intervention. Usual care comprised HIV testing, pre-exposure prophylaxis (PrEP), referral to government antiretroviral therapy (ART) services, contraception, condoms, syndromic management of sexually transmitted infections, health education, legal advice, and peer support. AMETHIST added peer-led microplanning tailored to individuals' risk and participatory self-help groups. All cisgender women (aged >18 years) who had sold sex within the past 30 days and lived or worked within trial cluster areas were eligible. Intervention status was not masked to programme implementers but was masked to survey teams and laboratory staff. After 28 months, a respondent-driven sampling (RDS) survey was done in the female sex worker population around each clinic, which measured the primary outcome, the combined proportion of female sex workers in the surveyed population at risk of transmitting HIV (ie, were HIV positive, not virally suppressed, and not consistently using condoms) or at risk of acquiring HIV (ie, were HIV negative and not consistently using condoms or PrEP). We report prespecified analyses of the disaggregated proportions of female sex workers in the surveyed population at risk of either transmission or acquisition of HIV. Analyses were prespecified, RDS-weighted, and age-adjusted. This trial is registered with the Pan African Clinical Trials Registry, PACTR202007818077777. FINDINGS: The AMETHIST intervention was started on May 15, 2019, and data were collected from June 1, 2019, until Dec 13, 2021. The RDS survey was done from Oct 18 to Dec 13, 2021, with 2137 women included in the usual care group (11 clusters) and 2131 in the AMETHIST intervention group (11 clusters) after excluding survey seeds (n=132) and women with missing key data (n=44). 1973 (46·2%) of the 4268 female sex workers surveyed were living with HIV; of these, 863 (93·5%; RDS-adjusted) of 931 women in the intervention group and 927 (88·8%) of 1042 in the usual care group were virologically suppressed. 287 (22·4%) of 1200 HIV-negative women in the intervention group and 194 (15·7%) of 1096 in the usual care group reported currently taking PrEP, of whom only two (0·4%) of 569 had protective tenofovir diphosphate concentrations in dried blood spots (>700 fmol/dried blood punch). There was no effect of the intervention on the primary endpoint of risk of both HIV transmission and acquisition (intervention group n=1156/2131, RDS-adjusted proportion 55·3%; usual care group n=1104/2137, RDS-adjusted proportion 52·7%; age-adjusted risk difference -0·9%, 95% CI -5·7% to 3·9%, p=0·70). For the secondary outcomes, the proportion of women living with HIV at risk of transmission was low and significantly reduced in the intervention group (n=63/931, RDS-adjusted proportion 5·8%) compared with the usual care group (103/1041, 10·4%), with an age-adjusted risk difference of -5·5% (95% CI -8·2% to -2·9%, p=0·0003). Risk of acquisition among HIV-negative women was similar in the intervention (n=1093/1200, RDS-adjusted proportion 92·1%) and the usual care group (1001/1096, 92·2%), with an age-adjusted risk difference of -0·6% (95% CI -4·6 to 3·4, p=0·74). INTERPRETATION: There was no overall benefit of the intervention on combined risk of transmission or acquisition. Viral load suppression in women living with HIV was high and appeared to be further improved by AMETHIST, suggesting potential for impressive uptake and adherence to ART in vulnerable and mobile populations. Sustaining treatment and reinvigorating prevention remain crucial. FUNDING: The Wellcome Trust and the Bill & Melinda Gates Foundation. TRANSLATIONS: For the Shona and Ndebele translations of the abstract see Supplementary Materials section.