Telemedicine for medical abortion increases access to safe abortion but its use has not been described in a controlled trial. We aimed to investigate the effectiveness, adherence, safety, and acceptability of a modified telemedicine protocol for abortion compared with standard care in a low-resource setting. In this randomised, controlled, non-inferiority trial we recruited women seeking medical abortion at or before 9 gestational weeks at four public health clinics in South Africa. Participants were randomly allocated (1:1) by computer-generated blocks of varying sizes to telemedicine or standard care. The telemedicine group received asynchronous online abortion consultation and instruction, self-assessed gestational duration, and had a uterine palpation as a safety measure. Participants in this group took 200 mg mifepristone and 800 μg misoprostol at home. The standard care group received in-person consultation and instruction together with an ultrasound, took 200 mg mifepristone in clinic and 800 μg misoprostol at home. Our primary outcome was complete abortion after initial treatment, assessed at a 6-week interview. Our non-inferiority margin was 4%. Group differences were assessed by modified intention-to-treat (mITT) analysis and per protocol. The trial is registered at ClinicalTrials.gov, NCT04336358, and the Pan African Clinical Trials Registry, PACTR202004661941593. Between Feb 28, 2020, and Oct 5, 2021, we enrolled 900 women, 153 (17·0%) of whom were discontinued before the abortion and were not included in the analysis. By mITT analysis, 355 (95·4%) of 372 women in the telemedicine group had a complete abortion compared with 338 (96·6%) of 350 in the standard care group (odds ratio 0·74 [95% CI 0·35 to 1·57]). The risk difference was -1·1% (-4·0 to 1·7). Among women who completed treatment as allocated (per protocol), 327 (95·6%) of 342 women in telemedicine group had complete abortion, compared with 338 (96·6%) of 350 in the standard care group (0·77 [0·36 to 1·68]), with a risk difference of -1·0% (-3·8 to 1·9). One participant (in the telemedicine group) had a ruptured ectopic pregnancy, and a further four participants were admitted to hospital (two in each group), of whom two had blood transfusions (one in each group). Asynchronous online consultation and instruction for medical abortion and home self-medication, with uterine palpation as the only in-person component, was non-inferior to standard care with respect to rates of complete abortion, and did not affect safety, adherence, or satisfaction. Grand Challenges Canada and the Swedish Research Council.