# The Cats.txt SEO Hoax That Fooled AI (And What It Reveals About LLMs)

Page: https://stenobird.com/podcast/the-edward-show-6568456/the-cats-txt-seo-hoax-that-fooled-ai-and-what-it-reveals-about-llms
Text version: https://stenobird.com/podcast/the-edward-show-6568456/the-cats-txt-seo-hoax-that-fooled-ai-and-what-it-reveals-about-llms.md
Podcast: [The Edward Show](https://stenobird.com/podcast/the-edward-show-6568456)
Published: 2026-05-31T12:48:00+00:00
Episode link: https://9342edaa-17ad-4315-a037-ec4dfabd63f7.libsyn.com/the-catstxt-seo-hoax-that-fooled-ai-and-what-it-reveals-about-llms
Audio file: https://traffic.libsyn.com/secure/9342edaa-17ad-4315-a037-ec4dfabd63f7/The_Cats.txt_SEO_Hoax_That_Fooled_AI_And_What_It_Reveals_About_LLMs.txt.mp3?dest-id=4081527
Processing state: not_requested
JSON: https://stenobird.com/v1/public/podcasts/the-edward-show-6568456/episodes/the-cats-txt-seo-hoax-that-fooled-ai-and-what-it-reveals-about-llms
Duration seconds: 466

## Resource

E1061: Breaking down one of the most interesting SEO experiments in recent memory: the Cats.txt hoax created by Mark Williams-Cook. He invented a completely fake "standard" called cats.txt, published formal documentation for it, and made it look legitimate. Soon after, major crawlers were requesting the file. Google indexed it. AI overviews described it as real. ChatGPT even said it could help you rank in search and large language models. Then the experiment went viral. Now AI systems acknowledge that it started as a joke. But before that happened, they confidently explained how it worked and why it mattered. This episode covers: - What the Cats.txt experiment actually was - How Googlebot, GPTBot, ClaudeBot, PerplexityBot, BingBot, AppleBot and others crawled it - Why ChatGPT initially claimed Cats.txt could help with rankings - What this reveals about how LLMs retrieve and synthesize information - Why "LLMs.txt" style tactics are often misunderstood - How consensus-looking content becomes treated as truth - The circular authority problem in AI systems - Why you can't reliably ask an LLM how its own infrastructure works - How this connects to previous experiments with fake schema markup The key takeaway: large language models do not inherently know what is authoritative. If enough content presents something as real, the model may confidently describe it as real. LLMs are very good at modeling what people say is true. That is not the same as knowing what is true. I also explain why you do not need fancy technical files like LLMs.txt to show up in AI-driven systems. A clear About page, strong positioning, relevant landing pages, brand mentions, and real marketing fundamentals will do more for you than trying to implement something that sounds advanced. If you are a busin…

## Actions

- request_transcript: `POST https://stenobird.com/v1/public/podcasts/the-edward-show-6568456/episodes/the-cats-txt-seo-hoax-that-fooled-ai-and-what-it-reveals-about-llms/transcription-requests` — Idempotently request low-priority transcript generation for this episode.
- read_markdown: `GET https://stenobird.com/podcast/the-edward-show-6568456/the-cats-txt-seo-hoax-that-fooled-ai-and-what-it-reveals-about-llms.md` — Read the agent-friendly Markdown representation of this episode resource.

A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed.

## Transcript

Full transcripts are not published on public pages unless there is a clear rights basis.