<p>Homophily—the tendency for individuals to associate with similar others—is a fundamental principle for understanding social network structure. In image-centric social networking services, visual content preferences have emerged as a salient dimension of social interaction, yet their structural role remains poorly understood. We define “visual homophily” as an operational measure of visual preference similarity computed over existing social ties, and characterize how it affects the structural properties of social networks. We construct user-level visual representations from posted images, quantify visual preference similarity between users, and integrate these scores as edge weights into the existing social network to define the Visual Interest Network. Structural analyses of Flickr data reveal that visual homophily is associated with systematic differences in structural properties known to influence diffusion—including a stronger intra-community concentration of edge weights and, under certain conditions, elevated similarity along bridging ties between semantically proximate communities. A permutation-based null-model analysis confirms that these structural patterns are statistically significant and not attributable to random weight assignment. The structural gain attributable to visual homophily is constrained by baseline network structure, while propensity score matching indicates that a significant association between visual homophily and diffusion outcomes remains after controlling for observed covariates in mature networks. These convergent findings are consistent with the interpretation that visual homophily may operate as a potential mediating factor within existing social relationships. By introducing a network-structural framework grounded in visual preference, this study advances understanding of how visual culture shapes online social network organization.</p>