BACKGROUND: HIV testing uptake in South Africa is below optimal levels. Community mobilization (CM) may increase and sustain demand for HIV testing, however, little rigorous evidence exists regarding the effect of CM interventions on HIV testing and the mechanisms of action. METHODS: We implemented a theory-driven CM intervention in 11 of 22 randomly-selected villages in rural Mpumalanga Province. Cross-sectional surveys including a community mobilization measure were conducted before (n = 1181) and after (n = 1175) a 2-year intervention (2012-2014). We assessed community-level intervention effects on reported HIV testing using multilevel logistic models. We used structural equation models to explore individual-level effects, specifically whether intervention assignment and individual intervention exposure were associated with HIV testing through community mobilization. RESULTS: Reported testing increased equally in both control and intervention sites: the intervention effect was null in primary analyses. However, the hypothesized pathway, CM, was associated with higher HIV testing in the intervention communities. Every standard deviation increase in village CM score was associated with increased odds of reported HIV testing in intervention village participants (odds ratio: 2.6, P = <0.001) but not control village participants (odds ratio: 1.2, P = 0.53). Structural equation models demonstrate that the intervention affected HIV testing uptake through the individual intervention exposure received and higher personal mobilization scores. CONCLUSIONS: There was no evidence of community-wide gains in HIV testing due to the intervention. However, a significant intervention effect on HIV testing was noted in residents who were personally exposed to the intervention and who evidenced higher community mobilization. Research is needed to understand whether CM interventions can be diffused within communities over time.