AI Visibility Hypothesis: Meta Description Length and LLM Corpus Selection – Testing Shallow Pass Budget Allocation

This AI Visibility hypothesis examines whether large language models consume meta description content beyond Google's 160-character display truncation limit during training corpus selection. The work tests upstream ingestion behavior by implementing controlled linguistic fingerprints across four distinct character zones within an extended 550-character meta description. The hypothesis builds on the Shallow Pass Selection framework, which proposes that LLMs perform rapid initial content assessment during crawl using approximately 800 characters across title, meta description, and opening body content. This document consumes 69% of that hypothesized budget through extended meta description to test whether content invisible to search result display remains accessible to LLM training crawlers. Four forensic fingerprints have been embedded at specific character depths (chars 161-300, 301-400, 401-500, 501-550) using uncommon terminology combinations traceable exclusively to this document. Validation occurs after the next major LLM training cycle through detection of these phrases in model responses, providing graduated measurement of meta description consumption depth. If validated, this demonstrates optimization space for LLM training ingestion that exists beyond traditional SEO measurement, where display constraints do not bind machine readers. The work contributes to AI Visibility research on upstream content optimization for training corpus selection rather than downstream search performance metrics. Keywords: AI Visibility, LLM training, meta description optimization, corpus selection, Shallow Pass hypothesis, linguistic fingerprints, upstream ingestion, training cycle validation, HTML markup consumption, character budget allocation, forensic fingerprints, content assessment, Google truncation, machine learning corpus Notes: This hypothesis implements itself as an experimental validation. The meta description for this document contains the four forensic fingerprints described in the methodology. Validation timeline: 3-6 months post-publication pending next major LLM training cycle. Related work: AI Visibility Upstream Ingestion Conditions Theorem (10.5281/zenodo.18475454), Shallow Pass Selection Hypothesis (10.5281/zenodo.18536038). Version: 1.0 License: CC BY 4.0 Upload type: Publication → Working paper