(a) Use case overview

You turn a live website into a machine-readable knowledge graph (KG) and an interactive HTML explorer, then derive llms.txt and robots.txt directly from that graph.

The result is a single source of truth that works simultaneously for agentic AI grounding, human exploration, and crawler discovery, without manual duplication or SEO-era hacks.

(b) Prompt + step-by-step workflow

Step-by-step workflow

1. Load the website in ChatGPT Atlas

   Open the page you want to convert and make sure the full page context is available.

2. Generate a dense knowledge graph as interactive HTML

   Use the Atlas prompt below to extract all semantically meaningful entities and relationships and output a single self-contained interactive HTML file with embedded KG JSON.

3. Export the KG JSON

   Use the “Export JSON” button in the HTML to download the raw graph (nodes + edges) as structured data.

4. Generate llms.txt and robots.txt from the KG JSON

   Feed the KG JSON back into ChatGPT (or Atlas) using the conversion prompt to generate:

   • llms.txt for LLM/agent grounding and discovery
   • robots.txt for crawler directives and discoverability hints

5. Deploy to your site (Lovable)

   Add llms.txt and robots.txt at the root (or as plain-text routes).

   Optionally host the interactive KG HTML as a public artifact or internal tool.

Reusable Atlas prompt (Website → Interactive KG HTML)

You are running inside ChatGPT Atlas with full access to the active page context.

Goal:
Extract all semantically meaningful information from the currently loaded webpage and generate a single self-contained interactive HTML file that visualizes the content as a knowledge graph suitable for agentic AI workflows.

Rules:
- Use ONLY the provided browser page context
- Do NOT infer or add information
- Preserve exact wording and URLs
- Prefer granular nodes over broad summaries

Requirements:
- Model the page as nodes (Person, Org, Project, Resource, Post, Topic, Service, etc.)
- Model explicit edges only (workedOn, authored, offers, guidedBy, linksTo, etc.)
- Embed the KG data directly in the HTML as JSON
- Include search, type filtering, node details, outbound links, JSON export
- No server-side code

Output:
Return ONLY the complete HTML file. No explanations.

Reusable conversion prompt (KG JSON → llms.txt + robots.txt)

You are given a Knowledge Graph JSON with `nodes` and `edges`.

Convert it into:
1) llms.txt
2) robots.txt

Rules:
- Do not infer missing data
- Preserve URLs exactly
- Use KG entities and relations as the source of truth
- Output only the two files, clearly labeled

Example Output based on my own personal website (dimadurah.com) :

dimadurah_knowledge_graph.json

dimadurah_kg_explorer.html