DREAMS aims to reduce HIV incidence among adolescent girls and young women (AGYW) by tackling drivers of HIV risk including gender-based violence. We evaluate the impact of DREAMS on recent experiences of violence perpetuated by men against AGYW. AGYW cohorts were randomly selected from demographic platforms in South Africa (rural KwaZulu-Natal) and Kenya (Nairobi informal settlements and rural Gem sub-county). AGYW aged 13-22 years were enrolled in 2017 (Nairobi, KwaZulu-Natal) or 2018 (Gem), with annual follow-up to 2019. We described proportions of AGYW who self-reported experiences of violence perpetrated by males in the 12 months preceding the interview, overall and by form (physical, sexual, emotional). We investigated associations with DREAMS (invitation to participate during 2017-2018) through multivariable propensity score-adjusted logistic regression and estimated the causal effect of DREAMS on experiences of violence, under counter-factual scenarios in which all versus no AGYW were DREAMS beneficiaries. Among 852, 1018 and 1712 AGYW followed-up in 2019 in Nairobi, Gem and KZN, respectively, proportions reporting any violence in 2019 were higher in Nairobi (29%) than Gem (18%) and KwaZulu-Natal (19%). By sub-type, emotional and physical violence were more frequently reported than sexual violence. We found no evidence of an impact attributable to DREAMS on overall levels of violence, in any setting. Nor was there evidence of impact on sub-types of violence, with one exception: an increase in physical violence in Nairobi if all, versus no, AGYW were DREAMS beneficiaries (16% vs 11%; +5% difference [95% CI: +0.2%, +10.0%]). Experiences of gender-based violence were common among AGYW, especially in urban settings, and DREAMS had no measurable impact on reducing violence within three years of implementation. Violence prevention programming that reaches more men and the broader community, sustained for longer periods, may yield greater gains in violence reduction than AGYW-focused programming. Additionally, more investment in implementation research is needed to bridge trial-based study findings from efficacy to population-level effectiveness.