llms.txt - Does It Actually Help AI Visibility, or Is It SEO's Newest Myth

What llms.txt actually is

llms.txt is a Markdown file placed at a domain's root (e.g. example.com/llms.txt) that holds a short summary of the site plus links to its core content pages. The idea: language models work with a limited context window, and converting full HTML into clean text is a messy, imprecise process - so a clean, condensed file written specifically for AI consumption could skip that step.

The proposal was published by Jeremy Howard, co-founder of Answer.AI, on September 3, 2024, with a full specification at llmstxt.org. It's worth being clear: this is a community proposal, not an official standard like robots.txt or sitemap.xml that search engines are obligated to respect.

What Google actually said - not interpretation, direct quotes

On June 17, 2025, Google's John Mueller wrote on Bluesky: "no AI system currently uses llms.txt", comparing the idea to the old meta keywords tag - a feature SEOs embraced enthusiastically that search engines never seriously relied on (source: seroundtable.com).

The following month, in July 2025, Google's Gary Illyes stated publicly and explicitly that Google does not support the llms.txt protocol and has no plans to. At the same event, he also noted that ordinary, solid SEO - not a dedicated AI-only file - is what actually drives appearances in AI Overviews (source: searchengineland.com).

What about ChatGPT, Perplexity, and Claude - do they actually read the file?

Here the answer is less clear-cut, because none of these companies has issued an official statement either way. Some site owners who checked their server logs report that OpenAI's crawler does request the llms.txt file regularly (source: searchengineland.com) - but a request for the file is not proof that its content actually shapes an answer returned to a user. Anthropic does publish its own llms.txt for its developer documentation, but its official guidance for site owners points to robots.txt, not llms.txt, as the binding mechanism for controlling its crawlers.

The practical takeaway: there is currently no solid, confirmed evidence from any major AI provider showing a causal link between having an llms.txt file and a higher chance of being cited.

So is it still worth adding one?

Adding an llms.txt file doesn't hurt - it's not a load issue, it won't harm your SEO, and it takes under an hour for a site owner who already knows their content. But it's also not the investment that moves the needle. If you have a free hour to spend on AI visibility, it pays off far more to work on the content's actual structure - not a separate file that no model has confirmed it even reads.

What's actually proven to drive AI citations

A study by Kevin Indig published on Search Engine Land found that 44.2% of ChatGPT citations come from the first 30% of a document (source: searchengineland.com). In other words, AI models tend to cite from the early part of a page - regardless of whether an llms.txt file sits beside it or not. The direct answer and the citable fact need to live near the top of the page, not buried in the seventh paragraph.

And as we showed in our article on schema vs. content structure, technical tagging alone doesn't move the needle - what works is clear, factual content built on real sources. That's exactly what this article itself is trying to demonstrate: one precise, real citation beats one more file nobody has confirmed they're reading.