A satirical experiment involving a fabricated web standard called ‘cats.txt’ highlighted the insufficient evidence often cited to promote LLMs.txt and other AI search optimization strategies globally.
The experiment demonstrated that a nonsensical file about office cats could satisfy all four common criteria used to validate the effectiveness of LLMs.txt, according to its creator, Dave Smart.
Smart developed ‘cats.txt’ as a mock standard for declaring office cats on websites, then subjected it to the same tests often used to prove LLMs.txt’s utility. The file was successfully crawled by AI bots, indexed by Google, and large language models (LLMs) returned specific details from it. Additionally, ChatGPT reportedly endorsed its SEO benefits.
The findings suggest that the current threshold of evidence for LLMs.txt and similar geo-optimization (GEO) tactics is so low that even a fictitious file about felines can meet it, Smart said.
Despite claims of its importance, no major LLM provider, including OpenAI, Anthropic, or Google, has officially documented using LLMs.txt for search or content discovery, according to industry observations. Google’s Search Advocate John Mueller previously indicated that Google does not use LLMs.txt.
Furthermore, studies conducted by analytics firms like Ahrefs have indicated that most web crawlers largely disregard LLMs.txt files. These studies reportedly found no measurable citation advantage for websites implementing the file.
The experiment aims to challenge the marketing narratives surrounding certain AI-driven search optimization methods by exposing the lack of rigorous validation behind them.
Source: Search Engine Journal
