Study: GEO rewriting can shift AI election-answer exposure

A new benchmark, VoteGEO-Bench, tests how generative engine optimization (GEO) rewriting changes which webpages AI engines cite on election queries covering the 2024 U.S. presidential race and Japan’s 2026 House of Representatives election, per researchers Shiqi Zhang and Xiaokui Xiao. Rewriting without naming a target rarely increased that page’s exposure, while rewriting that named a target and reorganized or strengthened its argument did; the study also found that an AI engine’s baseline exposure gap between explicit partisan voter identities can exceed the gain from rewriting a single page.