Introduction Implementation data for digital unsupervised HIV self-testing (HIVST) are sparse. We evaluated the impact of an app-based, personalised, oral HIVST program offered by healthcare workers in Western Cape, South Africa. Methods In a quasirandomised study (n=3095), we recruited consenting adults with undiagnosed HIV infection from township clinics. To the HIVST arm participants (n=1535), we offered a choice of an offsite (home, office or kiosk based), unsupervised digital HIVST program (n=962), or an onsite, clinic-based, supervised digital HIVST program (n=573) with 24/7 linkages services. With propensity score analyses, we compared outcomes (ie, linkages, new HIV infections and test referrals) with conventional HIV testing (ConvHT) arm participants (n=1560), recruited randomly from geographically separated clinics. Results In both arms, participants were young (HIVST vs ConvHT) (mean age: 28.2 years vs 29.2 years), female (65.0% vs 76.0%) and had monthly income <3000 rand (80.8% vs 75%). Participants chose unsupervised HIVST (62.7%) versus supervised HIVST and reported multiple sex partners (10.88% vs 8.7%), exposure to sex workers (1.4% vs 0.2%) and fewer comorbidities (0.9% vs 1.9%). Almost all HIVST participants were linked (unsupervised HIVST (99.7%), supervised HIVST (99.8%) vs ConvHT (98.5%)) (adj RR 1.012; 95% CI 1.005 to 1.018) with new HIV infections: overall HIVST (9%); supervised HIVST (10.9%) and unsupervised HIVST (7.6%) versus ConvHT (6.79%) (adj RR 1.305; 95% CI 1.023 to 1.665); test referrals: 16.7% HIVST versus 3.1% ConvHT (adj RR 5.435; 95% CI 4.024 to 7.340). Conclusions Our flexible, personalised, app-based HIVST program, offered by healthcare workers, successfully linked almost all HIV self-testers, detected new infections and increased referrals to self-test. Data are relevant for digital HIVST initiatives worldwide.