We explore estimation methods of seismic tsunami hazard based on the generalized extreme value theory, provide two case studies for constructing seismic activity models: one based on the generalized extreme value distribution to build a seismic activity model for Ryukyu trench subduction zone, and the other based on the generalized Pareto distribution to build a seismic activity model for Manila trench subduction zone. Six specific sites along the southeast coast of China are selected, and strong seismic activity models are constructed using generalized extreme value theory, considering the Manila trench and Ryukyu trench subduction zones as potential source areas. Through numerical simulation of seismic tsunamis, the estimation result of seismic tsunami hazard for these specific sites of Manila trench and Ryukyu trench subduction zones in a future time period is obtained. These results are then compared with estimation results that consider the uncertainty effects. The findings indicate that the ranking of deterministic estimation results is consistent with that of coupled uncertainty effects, with the former having higher values than the latter, but still having reference value.