Extinction vulnerability in seabirds likely emerges from combinations of traits rather than single attributes. We construct a weighted phenotypic similarity network of 108 oceanic bird species, where nodes are species and edge weights count shared traits (body size, feeding mode, feet morphology, migration pattern, and flight/locomotion behavior). Louvain clustering reveals four phenotypic clusters; one is composed of nearly half threatened species, compared with \({\le }20\%\) in two of the clusters. We interpret cluster profiles using complementary views: cross-cluster enrichment ( \(P(\text {cluster}\mid \text {trait})\) ) and within-cluster composition ( \(P(\text {trait}\mid \text {cluster})\) ). The most vulnerable cluster is dominated by medium body sizes (crow- and mallard-sized) and by surface-feeding, palmate feet, and pelagic migration. Across the four clusters, sparrow-sized shows the strongest enrichment for this cluster. Benchmarking against Erdős–Rényi, configuration, and Watts–Strogatz null models indicates high clustering coefficients with short paths and a small-world architecture most consistent with a Watts–Strogatz network. These results show how a trait similarity network can reveal clustered vulnerability and offer screening signals for conservation prioritization.