VIEX — Visibility Exclusion Index: why search-seeded corpora underestimate algorithmic citation in generative AI engines

This study investigates the selection bias introduced by corpus construction in empirical research on algorithmic citation in generative search engines (GEO). GEO research has seeded its corpora from organic search results, a procedure that captures only the intersection of entities visible in search and entities cited by the models. Through named-entity recognition over 5,544 responses from three models (Gemini 2.5 Flash, GPT-4o, Claude Sonnet 4.6) across two verticals, the search-seeded corpus was found to exclude between 47.5% and 73.2% of the cited entities — including the two most-cited systems (100% and 99% of category responses). The Visibility Exclusion Index (VIEX) is proposed as a measure of bias magnitude and a condition of sampling validity for comparative claims in GEO research. Version 2 (erratum and revision). This version corrects and strengthens v1: Reference corrections: the page numbers of Aggarwal et al. (2024) were corrected (from 5126-5137 to 5-16); orphan references were removed; Telles DOIs were corrected; and the co-authorship of the CIAH 2021 reference was corrected (from "Xavier" to "Dionathan"). Copy-editing revision (with a word guardrail: only punctuation was adjusted). Declaration of AI use added, for methodological transparency. The central thesis (search-seeded corpora systematically underestimate algorithmic citation, and VIEX measures the magnitude of that exclusion) remains and is reinforced.