BACKGROUND: There are conflicting findings about whether mobile phone text message reminders impact on antiretroviral adherence. We hypothesized that text reminders sent when dosing was late would improve adherence and HIV viral suppression. METHODS: Antiretroviral therapy (ART)-naive participants, from a South African outpatient ART clinic, were randomized to standard of care (SoC, 3 pretreatment education sessions), or intervention (SoC and automated text reminders if dosing >30 minutes late). Dosing time was recorded by real-time electronic adherence monitoring devices, given to participants at ART start. CD4 cell count and HIV RNA were determined at baseline, 16 and 48 weeks. Primary outcome was cumulative adherence execution by electronic adherence monitoring device. HIV-1 viral suppression (<40 copies/mL) at week 48 and count of treatment interruptions (TIs) >72 hours were secondary outcomes. Analysis was by intention to treat (missing = failure). Registration was with the Pan-African Clinical Trials Registry: PACTR201311000641402. RESULTS: A total of 230 participants were randomly assigned to control (n = 115) or intervention (n = 115) arms. Median adherence was 82.1% (interquartile range, 56.6%-94.6%) in the intervention arm, compared with 80.4% (interquartile range, 52.8%-93.8%) for SoC [adjusted odds ratio for adherence 1.08; 95% confidence interval (CI): 0.77 to 1.52]. Suppressed HIV RNA (<40 copies/mL) occurred in 80 (69.6%) of control and 75 (65.2%) of intervention (adjusted odds ratio for virological failure in intervention arm 0.77; 95% CI: 0.42 to 1.40). In the intervention arm, the count of TIs of >72 hours was reduced (adjusted incident rate ratio, 0.84; 95% CI: 0.75 to 0.94). CONCLUSIONS: Text message reminders linked to late doses detected by real-time adherence monitoring reduced the number of prolonged TIs, but did not significantly improve adherence or viral suppression.