BACKGROUND: Improving HIV screening in key populations is a crucial strategy to achieve the goal of eliminating AIDS in 2030. Social networking platforms can be used to recruit high risk-taking men who have sex with men (MSM) to promote the delivery of voluntary counseling and testing (VCT) as mobile HIV testing. Therefore, client recruitment and availability of mobile HIV testing through social networking platforms requires further evaluation. OBJECTIVE: The aim of this study is to compare the effects of targeting high risk-taking MSM and HIV case finding between two mobile HIV testing recruitment approaches: through the traditional website-based approach and through social networking platforms. METHODS: A comparative study design and propensity score matching was applied. The traditional VCT model, that is, the control group, recruited MSM through a website, and a trained research assistant visited the walk-in testing station at a gay village on Friday and Saturday nights. The social networking VCT model, the experimental group, recruited MSM from social networking platforms by periodically reloading into and conducting web-based discussions on dating apps and Facebook. The participants then referred to others in their social networks via a popular messenger app in Taiwan. The test was conducted at a designated time and place during weekdays by a trained research assistant. Across both modes of contact, before the mobile HIV testing, participants needed to provide demographic characteristics and respond to a questionnaire about HIV risk-taking behaviors. RESULTS: We recruited 831 MSM over 6 months, with a completion rate of 8.56% (616/7200) in the traditional VCT model and 20.71% (215/1038) in the social networking VCT model. After propensity score matching, there were 215 MSM in each group (mean age 29.97, SD 7.61 years). The social networking model was more likely to reach MSM with HIV risk-taking behaviors, that is, those seeking sexual activity through social media, having multiple sexual partners and unprotected anal intercourse, having experience of recreational drug use, and never having or not regularly having an HIV test, compared with the traditional model. HIV positive rates (incidence rate ratio 3.40, 95% CI 1.089-10.584; P=.03) and clinic referral rates (incidence rate ratio 0.03, 95% CI 0.001-0.585; P=.006) were significantly higher among those in the social networking VCT model than in the traditional VCT model. CONCLUSIONS: Through effective recruitment strategies on social networking platforms, the social networking VCT mode can be smoothly promoted, as compared with the traditional VCT model, to target high risk-taking MSM and increase testing outcomes.