Recommendation algorithms are central to content consumption on modern social media platforms as they personalize user experience by continuously feeding tailored content and subtly guiding users toward structurally influential clusters of videos. This dynamic, often indistinct to the user, can disproportionately shape exposure and engagement patterns. In this study, we propose a hop-aware random walk framework to evaluate the discoverability of such attractors in a real world YouTube’s recommendation network. Our approach contrasts two traversal strategies namely Uniform Random Walks (URW), simulating neutral exploration, and Degree-Biased Random Walks (DBRW), which prioritize transitions toward high-degree nodes. To model realistic user navigation, we initiate walks from varying topological distances (1-hop to 3-hop) surrounding the target structures. We apply this approach across multiple social network analysis techniques, including community detection algorithms, centrality-based rankings, and Focal Structure Analysis (FSA). By measuring average first-hitting times, we assess how easily users encounter different structural groups. Our results show that FSA-identified structures are consistently more reachable, especially under degree-biased walks. These findings offer new insights into the interplay between network topology and content visibility, with implications for fairness, transparency, and algorithmic auditing in recommendation systems.