This site now serves dineshmannam.com/llms.txt. Hugo generates it from the content tree on every build, so a new post lands in it without me editing anything. That change shipped as build config, so it never showed up as a post. This is the record of it.
If you have not run into the convention: llms.txt is a markdown index of a site’s content, meant to hand a language model a clean map instead of making it parse a pile of HTML. Jeremy Howard of Answer.AI proposed it in September 2024, and the spec lives at llmstxt.org. A plain text file at the site root, an H1, a one-line summary, then sections of links with short descriptions. That is the entire idea.
What the evidence says
Nothing appears to read it.
Google is explicit. At a Search Central Deep Dive in July 2025, Gary Illyes said Google does not support llms.txt and is not planning to, and that ordinary SEO is all you need to show up in AI Overviews (Search Engine Land). John Mueller had put it more bluntly on Reddit that April: “AFAIK none of the AI services have said they’re using LLMs.TXT (and you can tell when you look at your server logs that they don’t even check for it). To me, it’s comparable to the keywords meta tag” (Search Engine Journal). The comparison stings because it is fair. Both are the site owner announcing what the site is about, when you could just go read the site.
No major provider has publicly committed to consuming the file in production, not OpenAI, Google, Anthropic, Meta, or Mistral. Scattered log reports show GPTBot fetching it, but a fetch is not a commitment, and nobody has said a model uses what is inside.
Server logs say the same thing. Limy, an AI visibility vendor reporting on its own monitoring, counted 408 requests for llms.txt across more than 500 million AI bot events in a 90-day window. Ahrefs checked 137,210 domains and found 97% of published llms.txt files got no requests in May 2026. SE Ranking, looking at nearly 300,000 domains, found 10.13% of them publish one. So a lot of sites are serving a file into a void, mine included.
The one exception
Coding assistants use it, on documentation sites. In the Ahrefs data, AI agents fetched these files more than AI search bots, 10.5% against 1.1%, with Anthropic’s Claude Code near the top. That is a real use case: a tool pulling a library’s current docs in so it stops guessing at an API.
Even there, someone has to hand the file over. Ahrefs found no AI bot requested one that was not there. LangChain’s mcpdoc is built for that, and you point it at a library’s llms.txt yourself. Cursor did not read them natively as of June 2025 (forum).
It is a documentation use case, not a personal blog one. No coding agent needs an index of my writing about money.
Why it is here anyway
One template, no ongoing maintenance. The file rebuilds itself, so there is nothing to keep in sync and nothing to go stale the next time I publish. If the convention gets adopted, the site already speaks it. If it does not, I am serving a few kilobytes nobody asks for.
I would rather make that bet than a claim. Adding this file has done nothing for traffic, ranking, or citations, and I have no reason to expect it will.