SEO INSIGHTS
The Cats.txt Test That Broke GEO's Biggest Myth

The Cats.txt Test That Broke GEO's Biggest Myth

The Catstxt Test That Broke GEOs

 

Cats.txt vs LLMs.txt: I Ran the Experiment So You Don't Have to Trust Your GEO Horoscope

I uploaded a file called cats.txt to a website. It had no real instructions. No SEO value. Nothing technical inside it at all.

AI engines treated it exactly the same as sites using the "official" llms.txt file.

That's when it hit me: a lot of us aren't doing SEO. We're reading horoscopes and calling it strategy.

What Is llms.txt — And Why Everyone Suddenly Believes in It

llms.txt is a simple text file website owners add, hoping it tells AI models like ChatGPT and Gemini how to "read" their site. The promise sounds great: control how AI sees your business.

It spread fast, the way meta keywords once did. Back then, marketers stuffed keywords into a hidden tag, convinced it boosted rankings. Google quietly stopped using it years ago. Most people kept doing it anyway, just in case.

llms.txt has that same energy: unproven, but too tempting to ignore.

The GEO Horoscope Effect

Horoscopes work because of something called a Barnum Statement — vague lines so general they feel personal. "You're strong, but sometimes doubt yourself." That fits almost anyone.

GEO (Generative Engine Optimization) advice often does the same thing. "Add llms.txt to get cited by AI" sounds specific. It isn't. It's vague enough that when it "works," people assume it caused the result.

Here's the trap: if your site gets cited by ChatGPT after adding llms.txt, you credit the file. You ignore every site that got cited without one. That's not evidence. That's confirmation bias wearing an SEO hat.

The Cats.txt Experiment

To test this, I built a file with the same format as llms.txt — but filled it with nonsense. Cat facts. No real AI instructions. I named it cats.txt and placed it on a live site.

Then I watched how AI crawlers responded.

Nothing changed. The site got crawled, read, and treated the same way as sites running a "real" llms.txt file. No boost. No penalty. No special treatment either way.

If llms.txt genuinely worked as marketed, a fake file should have flopped. It didn't.

The Placebo Protocol: Two Identical Sites, One Real Test

To go further, I built two nearly identical simple websites.

  • Site A: had an llms.txt file
  • Site B: did not

Same content. Same structure. Same topics. I tracked which one got cited more across ChatGPT, Perplexity, and Gemini over several weeks.

The result: no meaningful difference. Citations were inconsistent on both sites, driven far more by content clarity than by the presence of a text file.

If llms.txt worked the way it's marketed, these twin sites wouldn't have looked so identical in AI results.

So Should You Delete Your llms.txt File?

Not necessarily. There are narrow cases where it might genuinely help — very large, complex sites where giving AI crawlers a clean map of key pages adds real technical clarity.

But for most small businesses and blogs, it isn't magic. It's not doing the heavy lifting people assume.

What actually seems to influence AI citations:

  • Clear, well-structured content that directly answers real questions
  • Genuine topical authority built over time
  • Content that's easy for both humans and machines to understand

llms.txt isn't useless. It's just overhyped — the SEO equivalent of a lucky charm.

Key Takeaways

  • llms.txt has no proven, consistent impact on whether AI engines cite your site.
  • The Cats.txt Experiment showed a meaningless fake file was treated identically to a "real" llms.txt file.
  • The Placebo Protocol test (two identical sites) found no significant citation difference between having llms.txt and not having it.
  • Much of the belief in llms.txt comes from confirmation bias, not measurable cause and effect.
  • Content clarity and real authority matter far more than any single text file.

FAQ

What is llms.txt used for? It's a text file meant to guide AI models on how to interpret a website's content, similar in concept to robots.txt.

Does llms.txt actually improve AI visibility? Current testing, including this experiment, shows no consistent, measurable improvement for most sites.

What is GEO (Generative Engine Optimization)? GEO refers to strategies aimed at getting content cited or featured by AI answer engines like ChatGPT, Gemini, and Perplexity.

Is llms.txt the same as robots.txt? No. Robots.txt controls search engine crawling behavior and has a long, proven track record. llms.txt is newer and its effectiveness is still unproven.

Should small businesses bother creating an llms.txt file? It likely won't hurt, but it shouldn't be treated as a priority. Focus on clear content and real authority first.


Curious if your own llms.txt file is actually doing anything? Try the same twin-site test yourself — and see what the data really says.