AI Signal 124
cats.txt introduced as joke standard meets same evidential criteria as llms.txt
↗
A satirical cats.txt file was created and publicly adopted, demonstrating that the four common “proofs” used to promote llms.txt can be satisfied by a clearly absurd file.
The experiment shows that the current evidential bar for SEO and AI-discovery claims is extremely low, meaning engineers could be misled into adopting standards that have no real effect. It warns practitioners to demand stronger validation before relying on llms.txt or similar GEO tactics for search ranking or AI grounding.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
cats.txt is a plain-text file placed at a site’s root that lists fictional office cats, their breeds, job titles, and a mandatory PurrLevel score.
The file was crawled by AI bots, indexed by Google, repeated by LLMs, and endorsed by ChatGPT, satisfying the same four “proofs” (crawled, indexed, repeated, endorsed) used to claim llms.txt works.
The stunt highlights that SEO and GEO evidence is weak, urging engineers to seek more rigorous proof before adopting llms.txt or related standards.
THE READ
What the cluster adds up to.
The change introduced is the creation of a deliberately absurd web standard called cats.txt. The author wrote a specification, published it on a blog, and announced it on LinkedIn, encouraging sites to place the file at their domain root. The SEO community responded, with at least one practitioner adding cats.txt to his own site and a separate site (catstxt.org) emerging to host a cleaner version of the spec. This mirrors the way llms.txt has been promoted, but with a transparent joke premise.
Cats.txt passed the exact four “proofs” that the industry cites for llms.txt: it was fetched by AI bots, indexed by Google, reproduced by large language models, and explicitly endorsed by ChatGPT. The author deliberately used these same criteria to show that they can be satisfied by a file containing no functional information about a website’s content or relevance. The fact that the file was treated as evidence despite its absurdity underscores the fragility of the evidential framework.
For engineers, the takeaway is that the current validation signals for SEO-related AI files are insufficient to guarantee any real impact on search ranking or model behavior. Relying on llms.txt or similar GEO tactics without deeper verification could lead to wasted effort and false confidence. The cats.txt experiment serves as a cautionary example that low-cost compliance does not equate to functional benefit.
Adopting cats.txt itself costs virtually nothing: a simple text file and a brief specification upload. By contrast, implementing llms.txt may involve additional documentation, monitoring, and coordination with search-engine guidelines, yet the experiment suggests those investments may not yield measurable results. The low barrier to creating a compliant-looking file means the evidential bar can be gamed easily.
The demonstrated “proofs” stop being meaningful at the point where actual search algorithms or LLM training pipelines would need to consume the file. The article notes that no major LLM provider has documented using llms.txt, and Google’s John Mueller explicitly said no AI system currently reads llms.txt. Therefore, while cats.txt can pass superficial checks, it provides no real SEO advantage, and the same limitation likely applies to llms.txt until stronger, documented integration is shown.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER