Recommendation systems often guide users into content traps, which are clusters of repetitive and reinforcing content that sustain attention [6] but limit the diversity of exposure [7]. While prior work has examined structural and topical drivers of these traps, the role of persuasive factors embedded in content in shaping them remains underexplored. This study bridges persuasion theory, content analysis, and complex social network analysis to investigate how persuasive features in YouTube transcripts contribute to trap formation. We construct recommendation graphs, extract persuasive cues grounded in established theories, and measure topical uniformity across focal structures. Our results show that persuasive features persist at higher levels in high-uniformity groups, where they amplify user engagement, while low-uniformity groups exhibit weaker persuasion–engagement dynamics. These findings highlight persuasion as a critical mechanism reinforcing algorithmic traps and introduce a framework for integrating psychological theory with network science to better understand and evaluate recommendation systems.