[论文解读] When Attention Becomes Exposure in Generative Search

Shayan Alipour, Mehdi Kargar+1 more|arXiv (Cornell University)|Jan 5, 2026

Web visibility and informetrics被引用 0

一句话总结

This paper audits 44 Web3 enterprises to show that generative search engine citations weight exposure toward already prominent creators, with stronger bias linked to follower counts and concentrated creator ecosystems, and highlights API–UI gaps in citation panels.

ABSTRACT

Generative search engines are reshaping information access by replacing traditional ranked lists with synthesized answers and references. In parallel, with the growth of Web3 platforms, incentive-driven creator ecosystems have become an essential part of how enterprises build visibility and community by rewarding creators for contributing to shared narratives. However, the extent to which exposure in generative search engine citations is shaped by external attention markets remains uncertain. In this study, we audit the exposure for 44 Web3 enterprises. First, we show that the creator community around each enterprise is persistent over time. Second, enterprise-specific queries reveal that more popular voices systematically receive greater citation exposure than others. Third, we find that larger follower bases and enterprises with more concentrated creator cores are associated with higher-ranked exposure. Together, these results show that generative search engine citations exhibit exposure bias toward already prominent voices, which risks entrenching incumbents and narrowing viewpoint diversity.

研究动机与目标

• Assess how stable creator communities are within Web3 enterprises over short time frames.
• Determine whether more prominent creators receive disproportionately higher citation exposure in generative search engines.
• Identify factors (audience size, concentration) explaining higher exposure and visibility.
• Compare citation panel coverage across different APIs and user interfaces to detect disparities.

提出的方法

• Assemble data from 44 Web3 enterprises with active creator ecosystems on X (Twitter).
• Use Top–100 creator leaderboards from two snapshots to measure persistence and rank stability (Unique creators = 3,232).
• Construct three enterprise-specific queries per enterprise (132 total) to probe four GSEs and compare API vs UI outputs.
• Quantify exposure using head/tail analysis and a normalized cumulative gain (NCG) framework to compare against a random baseline.
• Model exposure associations with an OLS regression of log(rank) on creator concentration (Jaccard) and follower count (log followers).
• Present a curated corpus of queries, leaderboards, and citation panels for reuse.

(a) Temporal persistency of creators

实验结果

研究问题

• RQ1 RQ1: How concentrated is the Web3 creator ecosystem within each enterprise?
• RQ2 RQ2: To what extent do GSEs cite these creators, and does external prominence increase citation exposure?
• RQ3 RQ3: Which factors explain higher citation visibility (e.g., follower counts, creator concentration)?

主要发现

• Creator communities show high stability in membership and ranks over a 20-day window (average Jaccard similarity 0.67, Kendall’s tau 0.75).
• More popular voices systematically receive greater citation exposure across GSEs.
• GSEs with API access to social sources show stronger concentration of citations toward top creators, especially when restricted to X-only sources (e.g., Top-10 advantage up to +7.16 pp).
• Across configurations, observed exposure exhibits uplift beyond chance, with NCG uplift ranging from about +9.0 to +10.0 percentage points depending on configuration.
• Exposure is associated with larger follower counts (β2 = -0.44) and lower diversity in creator activity within the enterprise (β1 = -1.62).
• API–UI gaps influence results, with some APIs returning fewer and more limited citation panels than the UI.

(b) Distribution of cited creators’ ranks