This study explores how Vietnamese EFL university students engage in self-directed vocabulary learning through YouTube. Guided by Activity Theory, it examines how YouTube functions as a mediational tool in EFL students’ self-directed vocabulary learning and analyzes the key factors shaping students’ engagement with YouTube. A total of 154 journals, 22 semi-structured interviews, and four focus group discussions were used as data for the study. Thematical analysis of these triangulated data uncovers seven significant learning patterns: (1) strategic and intentional, (2) incidental and affectively engaging, (3) personalized and interest-driven, (4) integrated with multiple digital tools, (5) repetitive for reinforcement, (6) influenced by educators and influencers, and (7) spontaneous and exploratory. Generally, it was found that YouTube served not merely as a content platform but as a dynamic mediating artifact in learners’ evolving, self-regulated learning systems. This study emphasizes the efficacy of informal, self-directed vocabulary acquisition through platforms such as YouTube and contributes to discussions on digital literacy practices, learner agency, and integrated teaching frameworks.