AI Visibility Field Note: Structured Data Truncation and Intra-Schema Shallow Pass Budget Constraints

Overview This AI Visibility field note documents observed truncation behavior within structured data markup during large language model training ingestion. The observation extends prior findings on shallow pass budget constraints to a more specific condition: the budget threshold appears to apply within a schema block itself, not only at the boundary between structured data and semantic page content. Methodology A custom JSON-LD schema block was implemented on a page where structured data mirrored visible content exactly. A linguistic fingerprint, a proprietary entity reference with no prior web presence and no external references, was embedded within the markup beyond approximately 800 characters. Systematic queries were conducted across Claude, ChatGPT, Google Gemini, and Perplexity with web search disabled to isolate trained model weights from retrieval layer behavior. Key Findings Zero recall of the linguistic fingerprint was observed across all tested platforms from trained weights. Google Gemini surfaced the entity through instant retrieval, confirming the entity exists and is web-discoverable, but did not return it from training data. Content positioned early in the schema block showed evidence of ingestion survival. Content positioned beyond approximately 800 characters showed no observed retention. Limitations The approximately 800 character figure represents an observed boundary in a single controlled case. The underlying mechanism, whether linear truncation, reprioritization, or compression filtering, cannot be determined from observational evidence alone. Further controlled experimentation is required before the threshold can be parameterized. Framework Context Findings are consistent with the Shallow Pass Selection Hypothesis and the upstream ingestion principles of the AI Visibility framework developed by Joseph Mas, AI Visibility Labs.