AI-driven recommendation systems increasingly influence how people discover and consume online content. Understanding how these systems shape user navigation has therefore become an important research problem. YouTube's recommendation algorithm, which drives nearly 70% of watch time on the platform, acts as a continuous intermediary between users and the content they encounter. In this study, we propose Weighted Focal Structure Analysis (WFSA), an extension of the traditional FSA framework. WFSA introduces edge weights based on recommendation rank and depth to better reflect how the recommendation hierarchy influences the network structure. Using datasets drawn from the 2024 Taiwan presidential election and the 2025 U.S. tariff expansion, we construct weighted recommendation graphs and extract focal structures using both FSA and five WFSA variants. Our results show that WFSA consistently identifies smaller, more tightly connected, and structurally influential groups than FSA does. Network resiliency experiments show that removing WFSA focal structures causes greater network fragmentation and connectivity loss. Among the evaluated weighting schemes, WFSA4 (linear rank decay, inverse depth decay with count-weighted aggregation) most effectively captures structurally critical groups. These findings suggest that recommendation hierarchy plays an important role in shaping which groups of content become influential within recommendation networks.

PDF