The Practical Guide to the Five ThatWare AI Stack Files
Understand what each AI Stack file does, what information belongs inside it, how the files work together, and how to deploy and maintain them across your website.
The ThatWare AI Stack converts existing website knowledge into five coordinated machine-readable assets:
Each file solves a different problem. Together, they create a clearer machine-readable layer around your website for AI crawlers, language models, retrieval systems, enterprise search, RAG workflows, and intelligent agents.
Supporting line: Five files. Five responsibilities. One connected AI-readiness architecture.
Connected Machine Discovery
Rather than feeding LLMs unexplained URLs or unstructured scraping feeds, the five files establish entry guidance, curated context, entity hierarchy, retrieval records, and semantic graphs.
Know Which File Solves Which Problem
The files are not intended to perform the same job. ThatWare AI's existing architecture deliberately separates guidance, curated context, structured organizational identity, retrieval records, and semantic relationships into coordinated assets.
| File | Primary Role | Core Question |
|---|---|---|
| ai.txt | AI guidance and entry point | Where should machines start? |
| llms.txt | Curated LLM context | Which content matters most? |
| ai-manifesto.json | Entity, identity and governance | Who are we and what is authoritative? |
| vector-feed.xml | Retrieval and vector preparation | What content can be ingested and retrieved? |
| semantic-sitemap.xml | Semantic website mapping | How is our content connected? |
The AI Stack Complements the Web You Already Have
The AI Stack should not be treated as a replacement for established website infrastructure. It does not replace:
- • robots.txt
- • sitemap.xml
- • Schema markup
- • Canonical tags
- • Internal linking
- • HTML accessibility
- • Technical SEO
- • High-quality website content
ThatWare AI itself describes the files as a complementary AI-readiness layer rather than a replacement for traditional crawling, indexing, structured-data, and SEO practices.
Crawl → Index → Navigate
Identity → Priority → Meaning → Relationships → Retrieval
Five Files Answer Five Different Machine Questions
"Where should I start?"
Acts as the concise orientation layer.
"Which resources should I understand first?"
Curates important content and explains why it matters.
"Who is this organization and what does it represent?"
Defines the organization, entities, source hierarchy, and governance information.
"Which content records are available for retrieval?"
Structures source records for ingestion, chunking, embeddings, RAG, and related workflows.
"How does everything relate?"
Maps pages, topics, entities, classifications, and semantic relationships.
Guidance → Context → Identity → Retrieval → Relationships
That is the logic behind the five-file architecture.
ai.txt - Give AI Systems a Clear Starting Point
ai.txt is the foundational guidance layer within the ThatWare AI Stack. Its purpose is to provide a concise machine-readable reference describing the website and pointing toward important AI-facing resources.
ThatWare AI currently positions ai.txt as the entry point for identifying site identity, preferred sources, important sections, connected AI Stack assets, usage guidance, and version information.
What ai.txt Should Answer
- • What website is this?
- • Which organization owns it?
- • Which resources are most important?
- • Where are the other AI Stack files?
- • Are there important content-use or access instructions?
- • When was the file last updated?
Recommended ai.txt Information
- • Website Identity: Primary domain and organization.
- • Description: A concise factual explanation of what the website represents.
- • Priority Resources: Important sections such as Products, Services, Documentation, Research, Policies, Support.
- • AI Stack References: Links to llms.txt, ai-manifesto.json, vector-feed.xml, semantic-sitemap.xml.
- • Governance or Usage Guidance: Only where appropriate and supported by the organization's actual policies.
- • Version Information: File version and latest update date.
ai.txt is the core file included in the Freemium Plan.
# AI Guidance File
site: https://www.example.com
name: Example Company
description: Example Company provides enterprise data-management software.
priority_sources:
- https://www.example.com/products/
- https://www.example.com/docs/
- https://www.example.com/research/
resources:
- llms.txt: https://www.example.com/llms.txt
- ai-manifesto.json: https://www.example.com/ai-manifesto.json
- vector-feed.xml: https://www.example.com/vector-feed.xml
- semantic-sitemap.xml: https://www.example.com/semantic-sitemap.xml
version: 1.0
last_updated: 2026-10-09This is an illustrative structure, not a universal web specification.
- • Small websites that want a simple AI-readiness starting point.
- • Growing companies that want a machine-readable orientation layer.
- • AI Stack deployments that need one location connecting the other assets.
llms.txt - Tell Language Models Which Resources Matter Most
llms.txt is intended to provide a concise, structured overview of a website and its most important resources. Rather than giving a machine an unexplained list of URLs, it can present priority resources with enough context to understand what each page contains.
ThatWare AI describes llms.txt as a Markdown-based resource for curating important content. It also correctly notes that llms.txt remains a proposed convention and adoption can differ between AI systems.
What llms.txt Should Answer
- • What does this organization do?
- • Which products or services matter most?
- • Where is the primary documentation?
- • Which URLs should be treated as authoritative references?
- • What does each important URL contain?
Recommended llms.txt Structure
- • Organization Summary: A short, factual overview.
- • Products: Priority product pages.
- • Services: Important service resources.
- • Documentation: Technical or operational references.
- • Research: Primary research and authoritative knowledge.
- • Policies: Important governance information.
- • Optional Resources: Secondary content that may still be useful.
ThatWare AI currently includes both ai.txt and llms.txt within the Boost Plan.
# Example Company
Enterprise workflow software for operations and automation teams.
## Products
- [Workflow Platform](https://example.com/platform/)
Core product overview and capabilities.
## Documentation
- [Developer Documentation](https://example.com/docs/)
API, integration and implementation guidance.
## Research
- [Research Center](https://example.com/research/)
Original reports and technical studies.
## Policies
- [AI Policy](https://example.com/ai-policy/)
AI governance and usage information.ai.txt (Orientation): Provides high-level website and AI-resource guidance.
llms.txt (Curation): Describes the content that language models or retrieval systems may find most useful.
Simple distinction: ai.txt says where to begin. llms.txt explains what to read.
SaaS platforms, Documentation websites, Publishers, Knowledge bases, Research organizations, Large service websites, Content-heavy brands.
ai-manifesto.json - Define the Organization Behind the Content
A website can contain accurate information but still leave fundamental identity relationships unclear. ai-manifesto.json is designed to provide a structured representation of organization identity, brand entities, products, services, content categories, canonical sources, ownership, authorship, and governance.
ThatWare AI explicitly presents this as a ThatWare AI structured asset, not as a universally adopted internet standard.
What ai-manifesto.json Should Answer
- • Who owns this website?
- • Which entities belong to the organization?
- • Which domains are official?
- • Which products and services belong to the brand?
- • What is the preferred source hierarchy?
- • Who owns or authors important content?
- • Which governance rules apply?
Recommended Information
- • Organization: Name, type, description, canonical domain.
- • Entities: Products, services, people, brands, frameworks.
- • Official Properties: Websites and authorized public profiles.
- • Content Categories: Research, services, documentation, support, etc.
- • Canonical Sources: Which pages should be considered primary for particular information.
- • Authorship & Ownership: Who is responsible for relevant knowledge.
- • Governance: Approved usage or source policies.
- • Versioning: Creation and modification timestamps.
{
"organization": {
"name": "Example Company",
"type": "Organization",
"url": "https://example.com"
},
"entities": [
{
"name": "Example Platform",
"type": "Product"
},
{
"name": "Enterprise Automation",
"type": "Service"
}
],
"sourceHierarchy": {
"product": "https://example.com/platform/",
"documentation": "https://example.com/docs/"
},
"version": "1.0",
"lastModified": "2026-10-09"
}Illustrative ThatWare AI implementation example rather than a universal JSON specification.
Consider a company with 3 brands, 8 products, 4 regional domains, 50 leadership profiles, and hundreds of pages. Without an explicit entity model, relationships remain implicit. ai-manifesto.json gives those relationships a structured representation.
Multi-product companies, Enterprise brands, Multi-domain organizations, Complex content ecosystems, Organizations with strong governance requirements, Brands with substantial entity complexity.
vector-feed.xml - Prepare Content for Retrieval Workflows
A webpage is optimized primarily for reading and browsing. A retrieval system often needs something different: stable records, source identifiers, content segments, entity metadata, and freshness information.
vector-feed.xml is designed to organize website content into records that can support ingestion, chunking, embedding, enterprise search, semantic retrieval, and RAG workflows. It should be treated as a preparation and interoperability layer, not as a guarantee that any external AI provider will consume the feed.
What vector-feed.xml Should Answer
- • What content record is this?
- • Where did it come from?
- • What is its stable identifier?
- • Which topic does it belong to?
- • Which entities does it reference?
- • When was it updated?
- • What is its relationship to other content?
Recommended Information
- • Stable Content ID: Example: DOC-00421.
- • Canonical Source: Original webpage.
- • Title: Page or section title.
- • Content Type: Product, Article, Documentation, Research, FAQ.
- • Topic: Primary subject.
- • Entities: Relevant organizations, products, people, concepts.
- • Language: Example: en.
- • Publication / Modification Dates: Important for freshness.
- • Parent Relationship: Which larger page or content object the record belongs to.
<vector-feed>
<document id="DOC-001">
<canonical>
https://example.com/services/ai-readiness/
</canonical>
<title>
AI Readiness Services
</title>
<type>service</type>
<topics>
<topic>AI Readiness</topic>
<topic>Machine Readability</topic>
</topics>
<modified>2026-10-09</modified>
</document>
</vector-feed>llms.txt: Curates important human-understandable resources.
vector-feed.xml: Structures retrieval records and metadata for downstream processing.
Simple distinction: llms.txt helps identify what matters. vector-feed.xml helps prepare it for retrieval.
RAG systems, Enterprise search, Internal copilots, Knowledge assistants, Large documentation repositories, AI search applications, Vector-based retrieval projects.
ThatWare AI describes source traceability and freshness management as key benefits of this file.
semantic-sitemap.xml - Map Meaning, Not Just URLs
A traditional sitemap answers: What URLs exist? A semantic sitemap attempts to answer: What do those URLs represent and how are they connected?
ThatWare AI positions semantic-sitemap.xml as a layer for describing content classifications, topics, entities, hierarchy, and relationships while continuing to use the standard sitemap.xml for conventional URL discovery.
What semantic-sitemap.xml Should Answer
- • What type of page is this?
- • What topic does it belong to?
- • Which entity is it primarily about?
- • What is its parent?
- • Which other resources are related?
- • Which language or region does it serve?
- • Which source is authoritative?
Recommended Information
- • Canonical URL
- • Page classification
- • Topic cluster
- • Primary entity
- • Parent entity
- • Child resources
- • Related content
- • Language
- • Region
- • Publish date
- • Modified date
- • Priority designation
<semantic-sitemap>
<entity id="example-company">
<name>Example Company</name>
<type>Organization</type>
</entity>
<entity id="ai-readiness">
<name>AI Readiness</name>
<type>Service</type>
</entity>
<relationship>
<source>example-company</source>
<target>ai-readiness</target>
<type>provides</type>
</relationship>
</semantic-sitemap>sitemap.xml: Primarily URL discovery.
semantic-sitemap.xml: Primarily meaning and relationships.
Use Both: ThatWare AI specifically recommends that the semantic sitemap complement, rather than replace, the conventional XML sitemap.
One Website. Five Views of the Same Knowledge.
Consider this example: A company publishes a page for: Enterprise AI Readiness Consulting
Tells machines: This is an important service resource.
Explains: This page describes the company's AI Readiness Consulting service.
Establishes: AI Readiness Consulting belongs to Example Company and is part of its official service architecture.
Creates: A retrieval-ready record tied to the canonical source and relevant entities.
Maps: Example Company → provides → AI Readiness Consulting.
This is why the files are more useful as a coordinated architecture than as unrelated standalone files.
Choose the Right AI-Readiness Depth
You Have a Small Website
Start with: ai.txt
Create a clear machine-readable entry point.
You Have Important Resources to Curate
Add: llms.txt
Highlight primary services, products, documentation, research, and policies.
Your Brand Has Complex Entities
Add: ai-manifesto.json
Structure organizational identity, products, people, brands, and governance.
You Are Building Retrieval or RAG
Add: vector-feed.xml
Create traceable content records suitable for downstream retrieval preparation.
You Operate a Large or Complex Site
Add: semantic-sitemap.xml
Expose semantic relationships across the entire content architecture.
File Selection by Use Case
This is a practical prioritization guide, not a restriction on what any industry can deploy.
| Use Case | Highest-Priority Files |
|---|---|
| Small business | ai.txt, llms.txt |
| SaaS website | ai.txt, llms.txt, ai-manifesto.json |
| Developer documentation | llms.txt, vector-feed.xml, semantic-sitemap.xml |
| Publisher | llms.txt, vector-feed.xml, semantic-sitemap.xml |
| E-commerce | ai-manifesto.json, vector-feed.xml, semantic-sitemap.xml |
| Multi-brand enterprise | All five |
| RAG implementation | vector-feed.xml + supporting identity files |
| Large knowledge base | llms.txt, vector-feed.xml, semantic-sitemap.xml |
Keep AI Assets Public, Stable and Easy to Locate
The exact deployment architecture should remain consistent with the implementation generated by ThatWare AI. For the current AI Stack, the product page shows assets referenced at predictable website-level paths such as:
Good Deployment Principles
- ✓ Use HTTPS: Serve machine-readable assets securely.
- ✓ Use stable URLs: Avoid changing file locations unnecessarily.
- ✓ Return successful HTTP responses: Files should be publicly reachable where intended.
- ✓ Avoid conflicting duplicates: Do not maintain multiple public copies that contain contradictory information.
- ✓ Use canonical URLs inside files: Internal references should point to authoritative website resources.
- ✓ Keep file references connected: Where appropriate, files should reference or align with related AI Stack resources.
Never Put Sensitive Data in Public AI Files
Machine-Readable Does Not Mean Everything Should Be Public. Do not include:
The AI Stack should represent approved public or intentionally exposed organizational knowledge. Personal information should never be included without a valid reason and authorization.
Validate Every File Before Publishing
Accessibility
Syntax
Identity
Sources
Freshness
Governance
Five Files Are Only Useful When the Information Is Reliable
Treating the Files as Ranking Hacks
The AI Stack does not guarantee rankings, mentions, citations, or recommendations.
Generating and Forgetting
A stale file can create conflicting signals.
Listing Every URL as High Priority
Priority loses meaning when everything is prioritized.
Creating Unsupported Entity Relationships
Do not associate products, people, credentials, or claims that are not genuinely connected.
Publishing Different Facts Across Files
All five assets should describe one coherent organizational reality.
Replacing Traditional SEO Infrastructure
Keep conventional sitemaps, schema, crawl controls, internal linking, and SEO.
Assuming Every AI Platform Supports Every File
llms.txt remains a proposed convention, and ThatWare-specific assets should not be represented as universal standards.
Update the Stack When Your Website Changes
The files should be reviewed when important organizational information changes:
Product Launch
Update: ai-manifesto.json, llms.txt, semantic-sitemap.xml, relevant vector records.
New Service
Update service relationships and priority sources.
Website Migration
Revalidate every canonical URL.
Rebrand
Update organization and entity information across all five files.
Leadership Change
Update entity and authority references.
Major Content Update
Refresh relevant vector records and semantic mappings.
Removed Content
Remove obsolete references.
ThatWare AI's product workflow itself ends with deployment and regeneration as website content changes.
Five Steps From Existing Content to AI-Ready Assets
01. Analyze
Discover accessible content, pages, metadata, and important entities.
02. Structure
Organize information into: Sources, Topics, Entities, Hierarchies, Relationships.
03. Generate
Create the AI Stack assets included within the selected plan.
04. Review & Validate
Check: Coverage, Syntax, URLs, Entities, File health, Exclusions, Freshness.
05. Deploy & Regenerate
Publish to stable website locations and update them when the underlying site changes.
This mirrors the five-step workflow currently described by the ThatWare AI platform.
Start Small or Deploy the Complete Stack
The current ThatWare AI pricing architecture maps these plans to the same file progression.
Freemium
Best for: Beginning AI readiness.
Includes:
- • ai.txt
Boost
Best for: Organizations that want both AI guidance and curated LLM context.
Includes:
- • ai.txt
- • llms.txt
Hyper-Intelligence
Best for: Organizations requiring the complete standard AI Stack.
Includes:
- • ai.txt
- • llms.txt
- • ai-manifesto.json
- • vector-feed.xml
- • semantic-sitemap.xml
Bespoke Enterprise
Custom website allowances, implementation requirements, assets, integrations, and enterprise architecture.
Better Machine Readability Is Not Guaranteed AI Visibility
Deploying these files does not guarantee:
AI systems apply their own: Retrieval methods, Model behavior, Source policies, Authority evaluation, Relevance logic, and Product rules. ThatWare AI itself states that AI Stack assets improve organization and readiness, while actual visibility depends on factors including relevance, accessibility, content quality, authority, platform policies, and individual retrieval systems.
The right expectation"Make your knowledge easier to interpret. Do not pretend you control the interpreter."