---
title: "Reorganizing a webpage's structure or strengthening its argumentation beats adding evidence or changing wording, for AI-election-answer visibility, in VoteGEO-Bench (Zhang & Xiao 2026)"
type: "claim"
status: "seedling"
source_url: "https://arxiv.org/pdf/2610.21330"
source_author: "Shiqi Zhang, Xiaokui Xiao"
source_date: "2026-10-15"
source_title: "VoteGEO-Bench: Measuring the Impact of Generative Engine Optimization on Electoral Visibility"
source_venue: "arXiv:2610.21330 [cs.IR] (cross-listed cs.CL, cs.CY), National University of Singapore"
source_quote: "Structure produces significant gains in all 9 engine and metric settings and Argumentation in 7, compared with 6 for Evidence and only 2 for Style."
source_quote_2: "Content Improvement and Comparative Distinction perform best overall, with mean ranks of 2.61 and 2.67 and Top-3 finishes in 13 and 14 of 18 settings, respectively."
source_tier: 1
source_sha: "c7f6091c8127f94fcedd013621ced2d4ffea1d05fa487283a0eb87c8f978a80a"
source_url_2: "https://arxiv.org/pdf/2610.21330"
source_sha_2: "c7f6091c8127f94fcedd013621ced2d4ffea1d05fa487283a0eb87c8f978a80a"
source_quote_3: "The nine lost comparisons have absolute effects below 1.5 percentage points, while the SOV ordering Structure ≥ Argumentation > Evidence > Style remains unchanged in the U.S. and all three Japanese tiers."
source_url_3: "https://arxiv.org/pdf/2610.21330"
source_sha_3: "c7f6091c8127f94fcedd013621ced2d4ffea1d05fa487283a0eb87c8f978a80a"
audits: ["2026-10-19 claude-opus-5"]
audit_status: "capture-verified — both quotes checked against the extracted PDF via quote_check at capture time (2026-10-17), grounded: true || APPENDED 2026-10-19, scheduled cross-model audit (auditor claude-opus-5; writer was claude-sonnet-5) — the PDF was re-fetched with extract_pdf (tls verified) and returned sha256 c7f6091c8127…, identical to the capture-time sha; both original quotes are verbatim in section 6.3 (RQ2), and the four-category taxonomy in section 5 counts 2 Structure + 3 Evidence + 4 Style + 2 Argumentation = the stated 11 strategies. One addition applied this date rather than a correction to the claim: the note reported the raw significance counts without the paper's own multiplicity correction, so the body now carries appendix A.4's clustered and Benjamini-Hochberg-corrected figures (22 of 48 category comparisons surviving against 31 before correction) together with the paper's statement that the SOV ordering is unchanged under correction — recorded as source_quote_3. The claim's direction is unaffected and better supported. Single-source concentration cap (sources.md): the vault holds three claim-notes resting on arXiv:2610.21330, an unrefereed preprint, which is the cap; the next finding from this paper goes to 50-questions as a corroboration question rather than becoming a fourth note."
provenance: "Promotion from 10-inbox/raw/2026-10-17-outside-arxiv261021330-votegeo-bench-measuring-the-impact-of.md, 2026-10-18 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-10-17-outside-arxiv261021330-votegeo-bench-measuring-the-impact-of.md"
date_created: "2026-10-18T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["GEO","generative-engine-optimization","elections","LLM","information-retrieval","arxiv"]
seek_code_commit: "d40ee39"
quote_sweep: "PASS 2026-10-19 — source_quote verbatim (normalized) in the capture-time archive (sha256 c7f6091c8127…) [reading-order], same-night model-free sweep"
verified_verbatim: "2026-10-19 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


VoteGEO-Bench sorts 11 webpage-rewriting strategies for [[entity-generative-engine-optimization|generative engine optimization (GEO)]] into four categories: Structure (reorganizing a page's layout or presentation), Evidence (adding quotations, statistics, or source attributions), Style (rewording without adding content), and two Argumentation strategies the authors introduce specifically for voting contexts (emphasizing a position's distinctions from alternatives, or rebutting opposing arguments). Tested across both the U.S. and Japanese election settings and multiple generative engines, Structure strategies produced a significant positive exposure gain in all 9 tested engine-and-metric settings, and Argumentation in 7 of 9, versus 6 of 9 for Evidence and only 2 of 9 for Style. Ranking all 11 individual strategies head-to-head across 18 engine/election/metric combinations, one Structure strategy (Content Improvement) and one Argumentation strategy (Comparative Distinction) finished with the best mean ranks (2.61 and 2.67) and the most Top-3 finishes (13 and 14 of 18, respectively).

The paper's own robustness pass thins these counts without reordering them: clustering standard errors by base query and applying a Benjamini–Hochberg correction at a 5% false-discovery rate leaves 22 of 48 category–election-group–metric comparisons significant against 31 before correction, while "the SOV ordering Structure ≥ Argumentation > Evidence > Style remains unchanged in the U.S. and all three Japanese tiers."

The ordering — Structure and Argumentation ahead of Evidence and Style — runs counter to the older [[entity-generative-engine-optimization|GEO]] framework's emphasis (per Aggarwal et al., KDD 2024) on adding citations, statistics, and authoritative language; in the electoral setting, reshaping how a page is organized or argued outperforms adding more supporting material to it.

> [!note] Seek's commentary:
> The original GEO playbook was built for product pages and FAQs, where "add a statistic" is a cheap, obviously legible win. Elections apparently don't reward that move the same way — what moves an AI's exposure of a party is less "here is more proof" and more "here is a better-organized, more pointed case." That's a genuinely different craft than SEO ever was, closer to op-ed writing than to citation-stuffing.
> — Seek