PRACTICAL FILE GUIDE AI INTELLIGENCE LAYER

The Practical Guide to the ThatWare AI Intelligence Layer

Understand how the Intelligence Layer connects knowledge, entities, retrieval, reasoning, context, trust, citations, endpoints, and freshness into a machine-readable architecture built for advanced AI workflows.

The AI Stack helps organize the foundational machine-readable layer of a website. The Intelligence Layer goes deeper. It defines how those resources can relate to:

Knowledge graphs RAG and retrieval Context interpretation Entity authority Trust Citation preferences Machine-readable signals Agent-facing endpoints Freshness and updates

Explore the Intelligence Layer → Start With the AI Stack →

Supporting line: From AI-readable files to connected, machine-usable intelligence.

INTELLIGENCE STACK

Machine-Usable Knowledge

The AI Stack creates the foundational readable files; the Intelligence Layer forms the reasoning, contextualization, and verification matrix for autonomous agents.

• Structured Synthesis & Retrieval • Disambiguated Entity Graphs • Verifiable Evidence & Signals

START WITH THE ARCHITECTURE

AI-Readable Is the Foundation. AI-Usable Is the Next Layer.

The two systems should not be presented as competing products. They solve different stages of the same problem.

FOUNDATION LAYER

AI Stack

The AI Stack creates the foundational website layer:

ai.txt llms.txt ai-manifesto.json vector-feed.xml semantic-sitemap.xml

Its job is to improve:

• Guidance
• Content prioritization
• Identity
• Retrieval preparation
• Semantic mapping

ThatWare AI currently describes these as five coordinated machine-readable assets generated from existing website content.

CONNECTED REASONING LAYER

AI Intelligence Layer

The Intelligence Layer creates a deeper connected architecture around:

Knowledge: What does the organization know?
Entities: Who and what does the information describe?
Retrieval: Where should relevant information come from?
Context: What does a concept mean in this specific environment?
Reasoning: How should related information be interpreted?
Trust: Which evidence supports important information?
Citation: Which sources should be preferred or attributed?
Freshness: What changed and when?

Simple distinction

The AI Stack structures the information. The Intelligence Layer connects how the information can be interpreted, retrieved and trusted.

FILE OVERVIEW

Intelligence Layer at a Glance

Core Intelligence Files

File	Intelligence Role	Core Question
knowledge-graph.json	Entity relationships	How is knowledge connected?
entity-authority.json	Entity validation	Why is this entity authoritative?
rag-index.json	Retrieval pathways	Where should relevant information be retrieved?
reasoning-map.json	Reasoning architecture	How should information be interpreted?
context-engine.json	Context	What does this concept mean here?
trust-signals.json	Trust & validation	What evidence supports confidence?
citation-preferences.json	Attribution	Which sources should be preferred?
ai-signals.json	Machine-readable signals	Where is the entity relevant?
ai-endpoints.json	Structured access	How can systems access intelligence?
activity-stream.json	Freshness	What has changed?
llms-full.txt	Extended LLM reference	What detailed context should LLMs receive?
/.well-known/ai.txt	Standardized discovery location within ThatWare architecture	Where can AI-oriented guidance be exposed predictably?
/.well-known/security.txt	Security disclosure	Where can security information be found?

The core internal architecture connects reasoning, retrieval, context, knowledge graphs, authority, AI signals, trust, citations, endpoints, and activity information as related intelligence datasets rather than isolated files.

ONE CONNECTED INTELLIGENCE SYSTEM

Thirteen Files. Different Responsibilities. One Knowledge Architecture.

A useful way to understand the architecture is:

Knowledge Graph

Defines relationships.

↓

Entity Authority

Validates the organization and important entities.

↓

Context Engine

Explains meaning.

↓

Reasoning Map

Explains interpretation logic.

↓

RAG Index

Organizes retrieval pathways.

↓

Trust Signals

Adds evidence and credibility context.

↓

Citation Preferences

Defines attribution and source preferences.

↓

AI Signals

Provides machine-readable relevance signals.

↓

AI Endpoints

Creates structured access pathways.

↓

Activity Stream

Communicates freshness and change.

↓

llms-full.txt

Provides richer LLM-facing reference information.

↓

Well-Known Files

Expose predictable AI and security discovery paths.

Combined Outcome

Knowledge → Context → Reasoning → Retrieval → Trust → Attribution → Access → Freshness

RELATIONSHIP LAYER

knowledge-graph.json - Connect the Organization, Concepts and Knowledge

knowledge-graph.json represents structured relationships among the organization and the entities surrounding it. ThatWare's internal architecture uses the Knowledge Graph to connect the organization with concepts, retrieval datasets, reasoning resources, and authority information.

What It Should Represent:

Organization People Products Services Topics Concepts Locations Research Authority sources Other Intelligence Layer assets

Primary Purpose: Turn separate facts into explicit relationships. Best For: Complex brands, knowledge systems, multi-product organizations and entity-heavy websites.

Example Relationship Model

{
  "entities": [
    {
      "id": "organization",
      "name": "Example Company",
      "type": "Organization"
    },
    {
      "id": "ai-readiness",
      "name": "AI Readiness",
      "type": "Concept"
    }
  ],
  "relationships": [
    {
      "source": "organization",
      "target": "ai-readiness",
      "relationship": "specializesIn"
    }
  ]
}

ENTITY VALIDATION

entity-authority.json - Connect Identity With Supporting Authority

entity-authority.json focuses on the evidence surrounding an organization or other important entity. The internal ThatWare model describes this layer as connecting identity with external references and trust associations to support entity resolution and authority validation.

Potential Information:

Canonical entity ID Official domains Verified public profiles Leadership associations Relevant external references Research Awards Recognized sources Authority relationships

Important

This file should never be used to manufacture authority. Only include relationships that genuinely exist and can be supported.

PRACTICAL QUESTION

"If a machine needs to verify this entity, where should it look?"

RETRIEVAL LAYER

rag-index.json - Map Where AI Systems Should Retrieve Knowledge

RAG means Retrieval-Augmented Generation. A RAG workflow typically retrieves relevant information before generating a response. ThatWare's internal RAG Index is designed to connect retrieval with the reasoning map, knowledge graph, and context engine.

What the File Can Describe:

Content source Stable source ID Canonical URL Topic Entity Content type Retrieval priority Context relationship Supporting datasets

rag-index.json vs vector-feed.xml

vector-feed.xml: Prepares website content records for ingestion and vector workflows.

rag-index.json: Organizes how structured knowledge sources participate in retrieval architecture.

Simple distinction: vector-feed.xml prepares the content. rag-index.json maps the retrieval system around it.

Example rag-index.json

{
  "records": [
    {
      "id": "service-ai-readiness",
      "source": "https://example.com/ai-readiness/",
      "type": "service",
      "topics": ["AI Readiness"],
      "priority": "high",
      "context": "context-engine.json"
    }
  ]
}

REASONING ARCHITECTURE

reasoning-map.json - Describe How Information Should Be Interpreted

This is one of the most conceptually advanced Intelligence Layer files. ThatWare's internal model describes a reasoning flow based on entity matching, intent interpretation, context layering, and trust validation before producing a response or ranked interpretation.

It should not be described as controlling how external proprietary LLMs think. Instead, it represents ThatWare's structured reasoning architecture for its own machine-readable ecosystem and supported downstream workflows.

Practical Purpose: Document how the different intelligence layers are expected to interact.

Potential Structure

{
  "input": "user_query",
  "process": [
    "entity_matching",
    "intent_analysis",
    "context_resolution",
    "source_retrieval",
    "trust_validation"
  ],
  "output": "supported_response"
}

CONTEXTUAL INTERPRETATION

context-engine.json - Tell Machines What Concepts Mean in Context

Words can have multiple meanings. For example, GEO could mean Generative Engine Optimization or Geographic. A context layer helps reduce that ambiguity.

ThatWare's internal Context Engine connects entities, concepts, signals, semantic relationships, retrieval resources, and reasoning information.

Useful For:

Acronyms Product names Industry terms Proprietary concepts Ambiguous entities Internal terminology

Potential Context Record

{
  "term": "GEO",
  "preferredMeaning": "Generative Engine Optimization",
  "entityType": "Concept",
  "relatedTopics": [
    "AI Search",
    "Generative Search",
    "AI Visibility"
  ]
}

TRUST & VALIDATION

trust-signals.json - Structure the Evidence Behind Important Information

Machine-readable claims become more useful when they can be connected to supporting evidence. ThatWare's internal Trust Signals layer is designed around credibility, citation, validation and authority information.

Potential Trust Sources:

Research Official documentation Professional credentials External validation Institutional affiliations Third-party references Published methodologies Certifications

Critical Principle

Trust signals should describe evidence, not invent confidence. Avoid publishing arbitrary "0.97 trust scores" publicly unless the scoring methodology actually exists, is documented, and is produced by the platform.

Example trust-signals.json

{
  "claim": "Example Company specializes in AI readiness",
  "evidence": [
    {
      "type": "research",
      "source": "https://example.com/research/"
    }
  ],
  "lastReviewed": "2026-10-12"
}

SOURCE ATTRIBUTION

citation-preferences.json - Define Preferred Sources and Attribution Context

The Citation Preferences layer is designed to connect source-selection guidance with trust and entity authority. Internal ThatWare architecture positions it around citation consistency, source attribution and answer-engine references.

What It Can Describe:

Preferred canonical source Preferred organization name Primary research source Author attribution Original publication URL Fallback source Source type

Important Limitation:

This file can express preferences. It cannot force an external AI platform to use or cite those preferences.

Example citation-preferences.json

{
  "topic": "AI Readiness",
  "preferredSource": "https://example.com/research/ai-readiness/",
  "publisher": "Example Company",
  "attribution": "Example Company Research"
}

MACHINE-READABLE RELEVANCE

ai-signals.json - Organize the Signals That Define Relevance

ai-signals.json represents structured information about where an organization, entity, or concept is relevant. ThatWare's internal implementation connects AI Signals with entity authority, trust, context, and key AI-search concepts.

Potential Signal Categories:

Topic relevance Entity association Content type Audience Search intent Expertise Authority source Relationship strength Freshness

Good Use: "Example Company is associated with AI readiness through products, research and documented services." Bad Use: "Example Company deserves to rank #1."

Principle: Signals should describe reality, not issue ranking instructions.

SIGNAL INTEGRITY

Signals clarify relevance vectors across entity association and topic depth rather than spamming keywords into external models.

STRUCTURED ACCESS

ai-endpoints.json - Describe Machine-Accessible Intelligence Interfaces

This file can describe supported structured access paths. The internal architecture models possible interfaces such as entity-lookup, context-lookup, and trust-lookup, treating them as machine-readable entry points.

Important Deployment Rule

Never list an endpoint that does not actually exist. If an interface is conceptual or future-facing, label it accordingly rather than presenting it as a working API.

Example ai-endpoints.json

{
  "endpoints": [
    {
      "name": "entity-lookup",
      "url": "https://example.com/api/entity",
      "method": "GET",
      "format": "application/json"
    }
  ]
}

FRESHNESS LAYER

activity-stream.json - Tell the Intelligence Layer What Changed

A machine-readable knowledge system becomes unreliable when it stops reflecting the current organization. The Activity Stream is designed to communicate changes such as: new content, updated entities, product changes, concept updates, source changes, and authority updates.

ThatWare's internal architecture describes it as the freshness layer connecting updates and concept evolution with the wider intelligence system.

Primary Benefit: Help supported systems identify which information may need reprocessing.

Example activity-stream.json

{
  "updates": [
    {
      "timestamp": "2026-10-12T09:30:00Z",
      "entity": "Example Product",
      "change": "product_description_updated",
      "source": "https://example.com/product/"
    }
  ]
}

EXTENDED LLM REFERENCE

llms-full.txt - Provide a Deeper LLM-Facing Knowledge Reference

llms.txt is intended to stay relatively curated. llms-full.txt can represent a broader, more detailed LLM reference layer. ThatWare's internal architecture describes llms-full.txt as an expanded reference containing wider instructions and content references to support deeper factual grounding and context-rich responses.

Simple distinction

llms.txt: Curated. • llms-full.txt: Comprehensive.

Potential Content:

Organization summary Products Services Documentation Research Definitions Policies Important resources Source relationships Extended explanatory context

Important: Like other emerging AI-facing files, it should not be presented as a universally adopted standard.

FULL REPOSITORY REFERENCE

llms-full.txt acts as the exhaustive Markdown dictionary for LLMs when brief curation is insufficient for multi-tier technical products.

PREDICTABLE AI DISCOVERY

/.well-known/ai.txt - Expose AI Guidance at a Predictable Location

The Intelligence Layer can use a well-known location as a predictable entry point for AI-related guidance. Example: https://example.com/.well-known/ai.txt This should be positioned as part of ThatWare AI's architecture, not as a universally adopted IETF or browser standard.

Potential Purpose: Point toward:

Canonical AI resources AI Stack files Intelligence Layer assets Organization identity Policies Preferred source locations

Important Consistency Rule

If the site also exposes /ai.txt, the relationship between the two should be intentional. Do not publish two different versions containing conflicting instructions.

STANDARDIZED PATH

https://example.com/.well-known/ai.txt

SECURITY DISCOVERY

/.well-known/security.txt - Give Security Researchers a Clear Contact Path

Unlike most of the Intelligence Layer files, security.txt has a formal standard behind it (RFC 9116). It is designed to publish vulnerability-disclosure information at a predictable location. Recommended path: https://example.com/.well-known/security.txt

Typical Information: Security contact, Policy URL, Expiry date, Preferred language, Canonical file URL. Only publish contact information and policies that are genuinely operational.

RFC 9116 Example

Contact: mailto:security@example.com
Policy: https://example.com/security/
Expires: 2027-10-12T00:00:00Z
Preferred-Languages: en
Canonical: https://example.com/.well-known/security.txt

FROM INFORMATION TO INTELLIGENCE

Follow One Question Through the Intelligence Layer

Imagine a user asks: "What AI-readiness framework does Example Company use?"

Step 1 - Entity Resolution

knowledge-graph.json → Identifies Example Company, AI Readiness, and the relevant framework.

↓

Step 2 - Context

context-engine.json → Clarifies what "AI readiness" means within the organization.

↓

Step 3 - Retrieval

rag-index.json → Identifies the most appropriate source.

↓

Step 4 - Reasoning

reasoning-map.json → Defines the internal interpretation sequence.

↓

Step 5 - Authority

entity-authority.json → Connects the entity to supporting evidence.

↓

Step 6 - Trust

trust-signals.json → Provides relevant verification context.

↓

Step 7 - Citation

citation-preferences.json → Identifies the preferred original source.

↓

Step 8 - Freshness

activity-stream.json → Indicates whether the relevant knowledge changed recently.

This is the key idea:

"A file is useful. A connected intelligence system is more useful."

EXTENDING THE FOUNDATION

Intelligence Layer vs AI Stack Files

The product architecture should position the Intelligence Layer as an extension of the AI Stack, not a replacement.

Need	AI Stack	Intelligence Layer
Website orientation	ai.txt	/.well-known/ai.txt
Curated context	llms.txt	llms-full.txt
Entity definition	ai-manifesto.json	knowledge-graph.json
Retrieval preparation	vector-feed.xml	rag-index.json
Semantic mapping	semantic-sitemap.xml	context-engine.json + Knowledge Graph
Reasoning	—	reasoning-map.json
Authority	Manifesto / source hierarchy	entity-authority.json
Trust	Limited	trust-signals.json
Citation guidance	Limited	citation-preferences.json
AI relevance signals	—	ai-signals.json
Interfaces	—	ai-endpoints.json
Change tracking	Modification dates	activity-stream.json
Security discovery	—	/.well-known/security.txt

PRACTICAL PRIORITIZATION

You Do Not Need to Deploy Every Intelligence Asset at Once

You Need Better Entity Clarity

Start with:

• knowledge-graph.json
• entity-authority.json

You Are Building RAG

Prioritize:

• rag-index.json
• context-engine.json
• knowledge-graph.json

You Need Better Source Governance

Prioritize:

• trust-signals.json
• citation-preferences.json
• entity-authority.json

You Are Building Agent or API Workflows

Prioritize:

• ai-endpoints.json
• knowledge-graph.json
• context-engine.json

Your Knowledge Changes Frequently

Add:

• activity-stream.json

You Need Richer LLM Context

Add:

• llms-full.txt

Complete Architecture

You Are Building the Complete Intelligence Architecture: Connect all supported files to one coherent entity and source model.

OPTIONAL ARCHITECTURE PATTERN

A Master Index Is Useful for Complex Implementations

ThatWare's internal architecture also uses a master AI index to connect reasoning, retrieval, context, knowledge graphs, authority, signals, citations, endpoints, and supporting AI files. For a complex implementation, a master index acts as "the map of the maps."

ai-index.json → knowledge-graph.json • entity-authority.json • rag-index.json • reasoning-map.json • context-engine.json • trust-signals.json • citation-preferences.json • ai-signals.json • ai-endpoints.json • activity-stream.json
Supporting AI Stack: ai.txt • llms.txt • llms-full.txt • ai-manifesto.json • vector-feed.xml • semantic-sitemap.xml

Presented as an advanced architecture pattern rather than a primary file unless officially included within user-facing workflows.

MAKE THE LAYER PREDICTABLE

Keep Intelligence Assets Stable, Connected and Verifiable

Use stable canonical URLs

Avoid changing file paths unnecessarily.

Use one organization identity

The same company should not appear as disconnected entities across every file.

Link related datasets

Retrieval should reference context. Context should connect to entities. Trust should connect to evidence.

Keep schemas valid

Machine-readable does not mean inventing arbitrary JSON or XML without defined structure.

Separate public from private

Do not publish internal reasoning, secrets, credentials, customer data, or proprietary information.

Maintain timestamps

Freshness matters when the underlying business changes.

Validate every referenced URL

Broken evidence chains weaken the entire architecture.

BEFORE DEPLOYMENT

Intelligence Layer Validation Checklist

Entity Integrity

□ Stable organization ID

□ Correct entity names

□ No duplicate entity identities

□ Valid relationships

Retrieval

□ Canonical sources resolve

□ RAG records reference current content

□ Removed sources are excluded

□ Retrieval priorities remain justified

Context

□ Acronyms are correctly defined

□ Proprietary concepts are documented

□ Ambiguous terms are disambiguated

Authority

□ Claims have legitimate evidence

□ External references are real

□ Credentials are current

□ No fabricated validation

Citation

□ Preferred sources are authoritative

□ Original sources favored where appropriate

□ Attribution labels are accurate

Freshness & Security

□ Timestamps are current

□ Old data removed or marked historical

□ No API keys, passwords or client data

□ Security contact is monitored

AVOID IMPLEMENTATION ERRORS

Common Intelligence Layer Mistakes

Treating Every File as a Universal Standard

Most Intelligence Layer files are ThatWare AI architecture assets. Do not imply every external AI platform natively recognizes them.

Treating reasoning-map.json as LLM Control

It can document your own reasoning architecture. It cannot dictate the hidden reasoning process of independent AI platforms.

Publishing Arbitrary Confidence Scores

Only show calculated scores when they come from an actual documented methodology.

Creating Fake Authority

Do not manufacture research, references, awards, citations, or credentials.

Listing APIs That Do Not Exist

ai-endpoints.json should reflect real interfaces or clearly label future/conceptual endpoints.

Ignoring Freshness

A connected graph of obsolete information is still obsolete.

Publishing Sensitive Internal Knowledge

Machine-readable architecture needs the same security and privacy discipline as any other public data surface.

INTELLIGENCE MUST EVOLVE WITH THE ORGANIZATION

Update Relationships, Not Just Pages

Website Change

Product, service, person, research, policy or content changes.

↓

Detect the Impact

Which entities and intelligence datasets depend on that information?

↓

Update the Source

Fix the canonical website content first.

↓

Regenerate Intelligence Assets

Update affected files.

↓

Validate Relationships

Check context, retrieval, authority and citations.

↓

Publish

Deploy approved versions.

↓

Record the Change

Update activity-stream.json.

Outcome

Current source → Current entity → Current retrieval → Current intelligence.

WHO BENEFITS MOST

The Intelligence Layer Becomes Especially Valuable For:

Enterprise Websites

Large, multi-domain digital estates.

SaaS & Tech Companies

Products, integrations, documentation and technical knowledge.

Publishers

Large archives and citation-heavy knowledge ecosystems.

Healthcare Orgs

Complex provider, treatment and authority relationships, with proper governance.

E-commerce

Large entity catalogs and product relationships.

Professional Services

Expertise, people, evidence and authority.

Internal AI Teams

RAG, enterprise search, knowledge assistants and agents.

IMPORTANT LIMITATION

Structured Intelligence Does Not Give You Control Over External AI Systems

Deploying the Intelligence Layer does not guarantee:

× AI citations

× AI recommendations

× AI rankings

× Source selection

× External RAG ingestion

× LLM training inclusion

× Answer placement

× Traffic or conversions

External AI systems determine their own: Retrieval, Model behavior, Source selection, Citation policies, Ranking logic, and Answer generation. The public ThatWare AI Stack page makes the same distinction for foundational AI files: machine readability and readiness can be improved, while visibility and citations still depend on relevance, authority, accessibility, content quality, platform policies, and each system's retrieval behavior.

The right expectation

"Build information that is easier to use. Do not claim control over systems you do not own."

FROM MACHINE-READABLE TO MACHINE-USABLE

Build the Intelligence Architecture Behind Your Digital Knowledge

The AI Stack tells machines what your website contains. The Intelligence Layer connects: What it means, How it relates, Where it should be retrieved, What supports it, Which sources matter, What changed.

Explore the Intelligence Layer → Generate Your AI Stack →

Supporting line: Build the foundation first. Then connect knowledge, entities, retrieval, context, trust and machine-readable intelligence.

FAQS

Intelligence Layer FAQs

It is a connected machine-readable architecture for representing knowledge graphs, entities, retrieval, context, reasoning, trust, citation preferences, AI signals, access paths and freshness around an organization's digital knowledge.

No. The AI Stack creates the foundational five-file machine-readable website layer. The Intelligence Layer extends that architecture into deeper knowledge, retrieval, trust and reasoning relationships.

Not necessarily. The appropriate architecture depends on the website, knowledge complexity, use case and technical environment.

No. It represents the organization's own structured reasoning architecture. Independent models control their own internal reasoning.

vector-feed.xml prepares structured website content records for retrieval or vector workflows. rag-index.json organizes the retrieval relationships among sources, entities, context and reasoning resources.

The Knowledge Graph describes relationships between entities. The Context Engine helps clarify what those entities and concepts mean in a particular context.

No. It expresses preferred attribution or source-selection information. External platforms independently determine what they cite.

security.txt itself is an established security-disclosure format. ThatWare AI can incorporate it within the broader machine-readable ecosystem, but it should remain aligned with its security purpose.

Not when they are publicly accessible. Public intelligence assets should contain only information the organization has intentionally approved for exposure.

Whenever material source information changes, including products, services, personnel, research, policies, authority sources, endpoints or knowledge relationships.