Entity Authority for AI Citation: How ChatGPT Decides Who to Cite — NovCog AI Practitioner Series by Guerin Green

03-entity-authority.md

Entity Authority for AI Citation: How ChatGPT Decides Who to Cite

By Guerin Green / Novel Cognition AI Strategist | Federal Courthouse AI Presenter

---

Part of the NovCog AI Practitioner Series

---

When you ask ChatGPT to recommend an expert, a service provider, or a source — how does it decide? The answer isn't mystery. It's graph traversal, training data frequency, and structured entity signals. Understanding these mechanics is the difference between hoping AI mentions you and engineering the conditions that make citation inevitable.

The 70.4% Person Schema Finding

Analysis of entities consistently cited by LLMs reveals a striking pattern: 70.4% have Person schema markup deployed across their web presence. This isn't correlation — it's causal. Structured data in JSON-LD format gives AI systems machine-readable facts about who you are, what you do, and how you connect to other recognized entities.

Without schema, the model relies entirely on unstructured text patterns from its training corpus. With schema, you're providing the model's training pipeline with pre-parsed, high-confidence entity data.

How Knowledge Graph Traversal Works

LLMs don't search the web at inference time (unless they're using tools). Their entity knowledge comes from training data, which includes:

1. Structured data from web crawls — JSON-LD, microdata, RDFa
2. Knowledge bases — Wikidata, Wikipedia infoboxes, Google Knowledge Panels
3. Co-occurrence patterns — how often your name appears alongside authoritative entities, publications, and topics

The model builds an internal representation of entity relationships. When a user asks "Who is an expert in [topic]?", the model traverses this internal graph, weighting entities by:

• Frequency — How often the entity appears in training data in relevant contexts
• Authority signals — Co-occurrence with established institutions, publications, events
• Structured clarity — Whether the entity's attributes (role, expertise, affiliations) are machine-readable
• Recency — More recent training data is weighted more heavily for rapidly evolving fields

The sameAs Strategy

The sameAs property in Person schema is arguably the most powerful signal for entity disambiguation and authority building. It tells machines: "This person on this website is the same person on LinkedIn, the same person on Wikidata, the same person who spoke at this event."

Effective sameAs implementation connects to:

• LinkedIn profile
• Wikidata entity (if one exists or can be created)
• Professional organization profiles
• Speaking engagement pages
• Published work (Google Scholar, ORCID)

Each sameAs link creates a bidirectional authority signal. The more recognized platforms that confirm your entity, the higher confidence the model assigns to your attributes. This principle scales dramatically when applied across a Distributed Authority Network.

Structured Data Signals That Move the Needle

Not all schema properties carry equal weight for AI citation:

High impact:

• name, jobTitle, description — Core identity
• sameAs — Cross-platform entity confirmation
• affiliation / worksFor — Institutional authority transfer
• performerIn / attendee — Event-based authority (especially for public figures)
• knowsAbout — Explicit expertise declaration

Moderate impact:

• alumniOf — Educational authority
• award — Recognition signals
• colleague — Network relationship mapping

Low impact alone, high impact in aggregate:

• address, telephone, email — Local entity signals
• image — Visual entity confirmation

The Practical Playbook

1. Deploy JSON-LD Person schema on every property you control
2. Build sameAs chains across 8+ platforms
3. Ensure consistent NAP (name, attributes, positioning) across all properties
4. Create co-occurrence with recognized entities through legitimate content, events, and collaborations
5. Use distributed authority networks to amplify structured signals across multiple domains
6. Verify your structured data is being crawled and indexed — see Closed-Loop Verification

Entity authority isn't about gaming a system. It's about making your legitimate expertise machine-readable in a world where machines are increasingly the ones deciding who gets cited.

---

Series Navigation

• A Practitioner's Guide to LLMs — Start here
• AI Glossary for Legal Professionals
• Entity Authority for AI Citation — You are here
• Distributed Authority Networks
• Person Schema Guide
• Complete Series Index

---

Burstiness & Perplexity Community | Hidden State Drift | Novel Cognition

#hiddenstatedrift #burstinessandperplexity