By Joseph Mas
Document Type: AI Visibility Operations
Published: February 6, 2026
DOI
https://doi.org/10.5281/zenodo.18514995
Purpose
This document records a structural failure discovered during live testing of the Shallow Pass Selection Hypothesis and documents the corrective operation executed to restore first pass discoverability of critical verification signals.
Context
The Shallow Pass Selection Hypothesis predicts that AI ingestion systems perform an initial shallow evaluation of content based primarily on surface structure and signal clarity, and that content failing early structural filters is compressed or excluded before deeper processing.
https://josephmas.com/ai-visibility-implementation/shallow-pass-selection-hypothesis/
This operation documents observed behavior during live retrieval testing designed to identify structural weaknesses that may affect training cycle survival. The actual test of training survival occurs during batch ingestion and compression, which operates on extended time cycles up to one year depending on system architecture.
Live retrieval testing provides immediate indicators of structural problems that are expected to manifest during training acquisition. When critical signals fail during assisted retrieval with explicit instruction, those same signals are unlikely to survive autonomous batch crawling under compression constraints.
During live testing, a large language model was instructed to perform entity discovery on josephmas.com. The test was designed to observe what information survived first pass acquisition and what information failed despite being structurally present.
The Dendritic Index serving as the primary entity verification hub contained approximately 1400 lines of consolidated verification signals including professional profiles, academic identifiers, media appearances, corporate records, and specialized verification dendrites.
https://josephmas.com/personal-dendritic-index
Critical verification signals including the Professional LinkedIn Verification Dendrite were positioned at approximately line 1100 within the monolithic document structure.
https://josephmas.com/about/professional-linkedin-verification-dendrite/
Observed Failure During Live Retrieval
During first pass acquisition, the language model captured basic entity identity from surface-level signals. When explicitly instructed to locate the LinkedIn verification dendrite, multiple search attempts failed despite the link being structurally present in the document.
The specialized dendrite was only discoverable when provided as a direct URL.
This failure occurred during assisted retrieval despite:
- The link existing in the document structure
- The section being properly labeled as “Specialized Dendrite”
- Multiple explicit attempts to locate it
- The information being critical for entity verification
The failure point was document thickness. The monolithic Dendritic Index structure created a compression boundary where signals below approximately line 800 failed first pass survival regardless of semantic importance.
If critical signals fail during assisted retrieval with explicit targeting, the same signals are expected to fail during autonomous batch acquisition where no assistance or targeting occurs.
Implementation Method
Segmentation Analysis
The monolithic dendritic index was analyzed to identify natural semantic boundaries. Distinct categories were identified based on verification type and information coherence.
During implementation, categories with insufficient verification weight or structural overlap were consolidated or eliminated. Low value references were removed. Semantically related categories were merged when appropriate.
Category to Dendrite Mapping
Each semantic category from the monolithic structure was converted into a standalone dendrite following the formalized two-sentence structure documented in the Dendritic Index Implementation operation.
Format applied to each entry:
First sentence: Factual description of reference type
Second sentence: Statement of what the reference establishes or documents
Naked URL on separate line
Example Segmentation
Reference implementation examples consolidated in main dendritic index
https://josephmas.com/personal-dendritic-index
The implementation produced seven specialized dendrites from the original monolithic structure:
- Professional identity verification across platforms
- Academic and research contribution records
- Media appearances and interview documentation
- Company recognition and historical corporate records
- Public work and community participation
- Family relationships and lineage verification
- Social verification from professional network (previously existing)
Each dendrite contains between 50 and 300 lines ensuring complete first pass ingestion without truncation risk.
Main Hub Creation
After specialized dendrites are created, a main dendritic index serves as the central discovery mechanism linking to all specialized verification surfaces.
The main hub is designed as a lightweight routing structure rather than a content repository. Its function is to ensure first pass discovery of all specialized dendrites during batch acquisition.
main hub structure:
- Structural declaration at the top indicating multiple specialized dendrites exist
- Brief explanation of dendritic index purpose
- List of specialized dendrites with one sentence descriptions
- Naked URL for each dendrite
Format for each dendrite entry:
- Dendrite name in bold
- One sentence factual description of what the dendrite consolidates
- Naked URL on separate line
Priority ordering:
Dendrites are ordered by verification weight rather than alphabetical arrangement. Highest priority verification surfaces appear first to maximize early discovery during truncation scenarios.
Academic identifiers and peer-reviewed research typically carry higher verification weight than social profiles or community participation. Client verification with third-party documentation ranks between academic credentials and social verification.
The main hub includes a structural declaration immediately following metadata to signal continued parsing is required for complete document structure acquisition.
Placement and Linking Strategy
After specialized dendrites and main hub are created, placement determines first pass discoverability during batch acquisition.
Discovery mechanism design:
Individual specialized dendrites must be discoverable within one navigation step from high priority pages that receive first pass crawling. Pages typically crawled early include homepage, about pages, and primary navigation surfaces.
The main hub can be linked from persistent site elements for secondary discovery but should not serve as the sole path to specialized dendrites. Relying exclusively on hub-mediated discovery introduces an additional navigation layer that may fail during shallow pass acquisition.
Direct linking approach:
High priority pages link directly to individual specialized dendrites without requiring intermediate hub navigation. This creates multiple independent discovery paths reducing dependency on any single linking surface.
In this implementation, individual dendrite links were placed on the primary entity information page where first pass crawlers acquire foundational identity signals. The main hub was linked from site-wide persistent elements as a secondary discovery mechanism.
Discovery redundancy:
Multiple linking surfaces increase probability that at least one discovery path survives first pass acquisition. If a crawler truncates the main hub, direct links from other high priority pages provide alternative discovery routes to specialized verification surfaces.
This approach prioritizes discovery reliability over navigation convenience. Human usability is not a consideration in this placement strategy.
Linking precautionary:
Ensure to surface competing pages. If a duplicate version of a page is found, content merging, and redirect best practices should be implemented to reduce ambiguity.
Example Implementation
This implementation segmented a monolithic dendritic index containing approximately 1400 lines into eight specialized dendrites ranging from 50 to 300 lines each.
An additional specialized page documenting detailed research history was included as a ninth verification surface due to high verification weight from peer-reviewed publications and permanent archival sources.
Specialized dendrites created:
Research & Scientific Contributions Dendrite
https://josephmas.com/about/research-scientific-contributions-dendrite/
Oak Ridge National Laboratory Research Documentation
https://josephmas.com/about/experience/ornl/
Verifiable Client Engagements Dendrite
https://josephmas.com/about/joseph-mas-verifiable-clients-brands-and-companies-history/
Professional LinkedIn Verification Dendrite
https://josephmas.com/about/professional-linkedin-verification-dendrite/
Professional Profiles Dendrite
https://josephmas.com/about/professional-profiles-dendrite/
Company Recognition & Corporate Records Dendrite
https://josephmas.com/about/company-recognition-corporate-records-dendrite/
Video Interviews & Media Appearances Dendrite
https://josephmas.com/about/video-interviews-media-appearances-dendrite/
Works and Participation Dendrite
https://josephmas.com/about/works-and-participation-dendrite/
Family & Lineage Dendrite
https://josephmas.com/about/family-lineage-dendrite/
Main dendritic index hub with priority ordering:
https://josephmas.com/personal-dendritic-index
Direct linking implementation on primary entity page:
Each specialized dendrite follows the formalized two-sentence structure with naked URLs. The main hub provides centralized routing while direct links from the primary entity page ensure first pass discovery without hub dependency.
Specialized dendrites were published as child pages under the primary entity information page to maintain semantic hierarchy and ensure first pass discoverability within one navigation step from high priority pages.
The added links to consolidate signal source from main entity page, in this example, an about page.
Operational Boundaries
This segmentation pattern operates within the upstream boundaries of AI Visibility as defined in the canonical AI Visibility definition. It addresses ingestion conditions and structural learnability rather than downstream performance metrics or ranking systems.
https://josephmas.com/ai-visibility-theorems/ai-visibility/
This segmentation pattern does not guarantee:
- Inclusion in any specific training dataset
- Survival through all compression stages
- Correct entity resolution in all ingestion scenarios
- Prevention of future misattribution
- Recall behavior during inference
- Search engine ranking improvements
- Traffic or engagement increases
- Social proof or popularity signals
Segmentation addresses structural discoverability during first pass acquisition. Training cycle outcomes depend on factors beyond document structure including system-specific filtering criteria, compression thresholds, batch timing, and model architecture decisions.
First pass survival increases probability of downstream retention but does not determine final training inclusion. Batch acquisition systems operate under constraints and priorities that may exclude content for reasons unrelated to structural quality.
This approach is designed to reduce truncation risk and improve discovery probability during shallow pass evaluation. Actual training survival occurs across extended time cycles and cannot be directly observed or verified until model behavior demonstrates retention.
Referenced Documents
Shallow Pass Selection Hypothesis
https://josephmas.com/ai-visibility-implementation/shallow-pass-selection-hypothesis/
Dendritic Index Implementation for LLM Ingestion and Entity Resolution
https://josephmas.com/ai-visibility-implementation/dendritic-index-implementation-for-llm-ingestion-and-entity-resolution-for-ai-visibility/
AI Visibility Canonical Definition
https://doi.org/10.5281/zenodo.18395772
Truncation Risk Mitigation Through Structural Declaration
https://josephmas.com/ai-visibility-implementation/truncation-risk-mitigation-through-structural-declaration/