Machine-readable
This page explains what an llms.txt file is, what ours says, and how much it is actually worth. The short version: it is a cheap, useful courtesy to any machine reading this site, it is not a ranking factor, and nobody should sell it to you as one.
By Simon Young, Founder, Question.Marketing.
Last updated 2 September 2026
llms.txt is a plain text file at the root of a website that gives large language models a short, structured summary of what the site is about and which pages matter most.
It sits at /llms.txt, in the same way robots.txt sits at /robots.txt. The format is simple markdown: a heading with the site name, a blockquote summary, then grouped lists of links with one line of context each.
The idea, proposed by Jeremy Howard of Answer.AI in September 2024, is that a model reading a website has a limited context window and a lot of navigation, scripts and boilerplate to wade through. A curated file removes the guesswork about what the site is and which URLs carry the substance.
No. llms.txt is a community proposal, not a ratified standard, and no major AI provider has publicly committed to reading it as a ranking or retrieval input.
We say that plainly because the honest answer matters more than the sales pitch. Google has said publicly that it does not use llms.txt. OpenAI, Anthropic and Perplexity have not documented support for it either.
So why publish one? Because the cost is close to zero, the file is genuinely useful to any human or agent inspecting the site, and adoption of a convention usually starts before support does. It is a cheap option on a plausible future, not a substitute for the work that actually moves visibility.
If anyone tells you an llms.txt file will get you cited in AI answers, they are selling you the easy part. The work that moves the needle is question research, answer-first pages, entity consistency and third-party corroboration.
robots.txt controls access, sitemap.xml lists every URL for crawlers, and llms.txt explains meaning and priority to a reader that has to summarise you in one pass.
robots.txt is permission: which user agents may fetch what. Ours explicitly welcomes GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended and Bingbot, because blocking the crawlers and then complaining you are invisible in AI answers is a self-inflicted wound.
sitemap.xml is inventory: a complete machine list of canonical URLs with change frequency and priority, aimed at classic search crawlers.
llms.txt is editorial: a short, opinionated map of the site written for a model that will read a few thousand tokens and then answer a question about you. All three are complementary, and all three are live on this site.
Ours contains a one-paragraph description of the business, what we do, our published prices, and a grouped index of every core, industry and company page with a line of context each.
Prices are in the file deliberately. If a model is asked what AEO costs in the UK, we would rather it read our real numbers than infer a range from someone else's page.
The file is generated from the same route table that produces the sitemap, so it does not drift when pages are added or renamed.
Yes, if your site already has substantive pages worth pointing at. No, if publishing the file is the whole plan.
A good llms.txt takes an hour. It forces you to answer a useful question: which twenty pages on this site would I want an AI to read if it could only read twenty? If you cannot answer that, the file is not your problem.
Keep it short, keep the descriptions factual, avoid marketing adjectives, and include prices and locations if you are comfortable with them being quoted back at you. Update it when the site changes.
Open ChatGPT, Claude or Gemini and ask it the question one of your customers asked you last week. If you aren't in the answer, that's what we fix. It starts with a measured baseline, not a contract.