INTRODUCTION: Respondent-driven sampling has been an effective sampling strategy for HIV research in many settings, but has had limited success among some youth in the United States. We evaluated a modified RDS approach for sampling Black and Latinx sexual and gender minority youth (BLSGMY) and evaluates how lived experiences and social contexts of BLSGMY youth may impact traditional RDS assumptions. METHODS: RDS was implemented in three cities to engage BLSGMY in HIV prevention or care intervention trials. RDS was modified to include targeted seed recruitment from venues, internet, and health clinics, and provided options for electronic or paper coupons. Qualitative interviews were conducted among a sub-sample of RDS participants to explore their experiences with RDS. Interviews were coded using RDS assumptions as an analytic framework. RESULTS: Between August 2017 and October 2019, 405 participants were enrolled, 1,670 coupons were distributed, with 133 returned, yielding a 0.079 return rate. The maximum recruitment depth was 4 waves among seeds that propagated. Self-reported median network size was 5 (IQR 2-10) and reduced to 3 (IQR 1-5) when asked how many peers were seen in the past 30 days. Qualitative interviews (n=27) revealed that small social networks, peer trust, and targeted referral of peers with certain characteristics challenged network, random recruitment, and reciprocity assumptions of RDS. HIV stigma and research hesitancy were barriers to participation and peer referral. CONCLUSIONS: Small social networks and varying relationships with peers among BLSGMY challenge assumptions that underlie traditional RDS. Modified RDS approaches, including those that incorporate social media, may support recruitment for community-based research but may challenge assumptions of reciprocal relationships. Research hesitancy and situational barriers must be addressed in recruitment and study designs.