What is llms.txt and should your business have one?

If somebody has mentioned llms.txt to you — in a quote, in a forum, or in a slightly breathless LinkedIn post about "AI-optimising your website" — you deserve a straight answer. This page gives you one, and everything in it is traceable to a named primary source with the date it was checked.

That matters here more than on almost any other topic, because llms.txt is a subject where speculation gets repeated as fact. Plenty of articles will tell you ChatGPT, Claude and Perplexity "use llms.txt". None of those three companies has said so.

What is llms.txt?

llms.txt is a single Markdown file placed at the root of a website — yoursite.co.uk/llms.txt — containing a short, structured description of what the site is and links to its most important pages, written for large language models rather than for human visitors. It was proposed by Jeremy Howard, co-founder of Answer.AI, on 3 September 2024, and the proposal lives at llmstxt.org.

The thinking behind it is straightforward. An AI tool trying to understand your business has to fetch pages full of navigation, cookie banners and JavaScript and work out from all that what you actually do. A well-written llms.txt hands it the answer directly: this is the business, this is what it does, these are the areas it covers, these are the pages worth reading. Think of it as a briefing note left on the reception desk.

What the llms.txt specification actually says

The specification requires exactly one thing: an H1 heading naming the project or site. Everything else — the summary, the prose, the link lists — is optional. That is a much looser document than most articles about it imply.

Version 2 of the specification was published on 10 August 2026, described by its author as an update "based on what I learned from two years of adoption". As of the version current on 15 September 2026, the file is expected to contain, in this order:

• An optional byte-order mark.
• An H1 with the name of the project or site. This is the only required section.
• An optional blockquote with a short summary containing the key information needed to understand the rest of the file.
• Optional further detail, in ordinary Markdown, as long as it contains no headings.
• Zero or more H2-delimited "file lists", each a list of Markdown hyperlinks, optionally followed by a colon and a note about the link.

Version 2 also loosened the location rule: the file may now sit at /llms.txt or at any subpath such as /docs/llms.txt, each covering the URLs beneath its own path. For a business with one site, the root is still the only sensible place.

One thing the specification does not define at all is llms-full.txt. That is a separate convention adopted by some software documentation sites, where every page is concatenated into one very large Markdown file. It has nothing to do with a plumber or a wedding venue.

Does Google use llms.txt?

No. Google's own documentation on its AI features, last updated 10 December 2025, says in as many words: "You don't need to create new machine readable files, AI text files, or markup to appear in these features. There's also no special schema.org structured data that you need to add."

Two Google staff have been blunter still. On 17 June 2025, John Mueller wrote on Bluesky: "FWIW no AI system currently uses llms.txt." And in comments reported on 1 August 2025, Gary Illyes said: "In the greater scheme of things though it's very easy to draw a parallel between 1990's keywords meta tag and this, and we all know how... useful the keywords meta tag became, very fast."

There is one wrinkle people cite as proof Google has changed its mind. In May 2026 Google added a check for the presence of an llms.txt file to Chrome's Lighthouse tool, in a new "Agentic Browsing" category, on the reasoning that without one, agents may spend more time crawling a site to work out its structure. It targets AI agents and browser tools, not Google Search, and Google has been clear it is not a ranking signal. A tool checking whether a file exists is not a search engine using it.

Do ChatGPT, Claude, Copilot or Perplexity use llms.txt?

Not one of them has confirmed it. This is the single most valuable sentence on this page, so here it is in detail, checked against each company's own published documentation on 15 September 2026.

• OpenAI publishes a page listing its crawlers — GPTBot for model training, OAI-SearchBot for ChatGPT's search results, ChatGPT-User for actions a person asks ChatGPT to take, and OAI-AdsBot. It explains how to control each one through robots.txt. It makes no claim that any of them reads your llms.txt.
• Anthropic publishes the same sort of page for ClaudeBot, Claude-User and Claude-SearchBot. It does not mention llms.txt anywhere.
• Perplexity documents PerplexityBot, which indexes pages for its search results, and Perplexity-User, which fetches a page when a person asks about it. Again, no claim about llms.txt.
• Microsoft has, as far as could be established on 15 September 2026, published no statement either way about whether Bing or Copilot reads llms.txt. If you see a confident claim that it does, ask for the source.

There is a genuine reason this myth persists. OpenAI and Perplexity both publish an llms.txt for their own developer documentation, so that coding assistants can pull in their API docs. People see docs.perplexity.ai/llms.txt and conclude that Perplexity reads yours. Publishing one and consuming one are different things.

The measured evidence points the same way. Ahrefs published a study on 15 June 2026 covering 137,210 domains using its web analytics product, of which 28% published an llms.txt file. Looking at May 2026 server logs, 97% of those files received zero requests — nothing fetched them at all. Among the 3% that were fetched, SEO audit tools were the largest single category at 21.7% of requests; all AI bots of every kind together made up 19.5%, of which AI retrieval bots such as OAI-SearchBot and PerplexityBot were just 1.1%. Slackbot fetched llms.txt files more often than PerplexityBot did. And no AI bot ever requested an llms.txt from a site that did not have one: they do not go looking.

llms.txt vs robots.txt vs sitemap.xml vs schema — what each file actually does

Four different files get muddled together in conversations about AI, and only one of them is llms.txt. Here is what each genuinely does, and who genuinely acts on it.

The costliest confusion is between the last two rows. llms.txt is an invitation; the opt-outs are a fence. If you want to keep AI companies from training on your content, llms.txt does nothing at all — you want robots.txt, and you need to know that blocking training crawlers is separate from blocking the retrieval crawlers that let ChatGPT cite you.

How to create an llms.txt file

Fifteen minutes and a text editor. There is no code, no plugin required and nothing to maintain. Write it, upload it, forget it.

What to put in it

Follow the specification's shape: your business name as an H1, a blockquote summarising what you do and where, a few sentences of detail, then H2 sections holding lists of links with short notes. Write the summary as though explaining the business to a stranger who will never see your website. "Friendly, reliable service for all your needs" is useless to a machine and, as it happens, to a customer.

A worked example

Here is a complete, specification-shaped file for a fictional dog groomer. Copy the shape, change every word.

Two details people get wrong. Use full https:// addresses, not relative ones like /prices — a tool reading the file may have no idea which domain it came from. And do not put a heading inside the detail paragraph; the specification reserves headings for the H1 and the H2 sections.

For the same shape on a real site rather than a fictional one, this site publishes one at websiteclarity.co.uk/llms.txt. It is longer, but the structure is identical.

Uploading it on WordPress, Wix, Squarespace and Shopify

The genuine hurdle is not writing the file but getting it to the root of your domain, and that varies sharply by platform.

Check it afterwards by typing your address plus /llms.txt into a browser. It should display as readable text. If it downloads instead, your server is sending the wrong content type; if you get a 404, it is in the wrong folder.

Is llms.txt worth it?

On the evidence, the file itself is worth close to nothing, and the writing of it is worth quite a lot. That is the honest split, and it is why the answer is "yes, eventually, if it is easy" rather than either "no, it is snake oil" or "yes, it is essential".

Three arguments survive scrutiny:

• It costs almost nothing. Fifteen minutes, no ongoing cost, no maintenance, no subscription. Anything sold to you with a monthly fee attached is a warning sign.
• It cannot hurt. A text file sitting at the edge of your site cannot break a page, slow anything down or affect your ranking. Worst case, nothing reads it.
• The exercise is the real prize. Writing it forces you to produce one clear paragraph saying exactly what you do and where — the thing most small business websites conspicuously lack. That paragraph then improves your homepage, your service pages and your Google Business Profile description. Even if no machine ever reads the file, writing it makes your website better.

The argument that does not survive is the one you will see most often: that llms.txt is "how you get recommended by AI". Nobody has shown that. If it were true, the 3% of files that do get fetched would be visited by assistants, not by SEO audit tools and Slackbot.

What should a small business do first instead?

Four things, all free to do yourself, every one with documented effects that llms.txt does not have. Do them in this order and llms.txt becomes the last item on a list you have almost finished.

• Your Google Business Profile. For a local service business this is usually the single largest source of enquiries, and it feeds the map pack that sits above the ordinary results. How to set up a Google Business Profile.
• A real page for every service, naming the places you cover. AI assistants answer from the text on your pages. If your site says "all your needs" where it should say "boiler repairs in Wrexham and Chester", there is nothing for anything to quote. How to write about your services.
• Schema markup. The structured data Google and Bing genuinely do parse, labelling your address, hours, service area and reviews as facts. Unlike llms.txt, it is a mature standard with documented uses. What schema markup is and whether you need it.
• Bing Webmaster Tools, with IndexNow switched on. Bing supplies web results to ChatGPT and Copilot, so it matters far more for AI search than its search share suggests, and it is the one place you can see AI citation data for your own site. Setting up Bing Webmaster Tools and enabling IndexNow.

If your interest in llms.txt came from wanting AI tools to recommend you, the guide on getting your business on ChatGPT covers what actually influences that. Short version: it is your pages, your profile and what other sites say about you — not a file at your root.

Common questions

What is llms.txt in simple terms?

It is a Markdown file you put at yoursite.co.uk/llms.txt, written for AI tools rather than for people. It names your business, summarises what you do and where, and lists links to your important pages, so a machine reading it gets the essentials without crawling your whole site. It is a voluntary proposal published by Jeremy Howard in September 2024, not a rule anyone has to follow, and no major AI assistant has confirmed reading it.

Is llms.txt an official standard?

No. It is a proposal published at llmstxt.org by Jeremy Howard of Answer.AI on 3 September 2024, updated to a version 2 on 10 August 2026. It has not been through the IETF, the W3C or any other standards body, and it has no governing organisation behind it. Robots.txt, by contrast, was formalised as RFC 9309 in 2022. Calling llms.txt a standard is the most common error in articles about it.

What is the difference between llms.txt and llms-full.txt?

llms.txt is the short index: a heading, a summary and lists of links. llms-full.txt is a convention some software documentation sites adopted separately, where the entire documentation set is concatenated into one enormous Markdown file so an AI tool can swallow it in one request. The llms.txt specification itself does not define llms-full.txt. A local business has no use for the full version.

Does llms.txt stop AI companies using my content?

No, and this is a dangerous misunderstanding. llms.txt is an invitation, not a fence. Blocking or permitting AI crawlers is done in robots.txt, using the crawler names each company publishes: GPTBot and OAI-SearchBot for OpenAI, ClaudeBot and Claude-SearchBot for Anthropic, PerplexityBot for Perplexity, and Google-Extended for Gemini training. Adding llms.txt controls nothing at all.

Will llms.txt get my business recommended by ChatGPT?

There is no evidence that it will. OpenAI has never said its assistants read your llms.txt, and Ahrefs found that in May 2026 AI retrieval bots such as OAI-SearchBot and PerplexityBot accounted for 1.1% of requests to the llms.txt files that were fetched at all. What AI assistants demonstrably do read is your ordinary pages, your Google Business Profile and what other sites say about you.

Do I need an llms.txt generator or a plugin?

No. The file is a few hundred words of Markdown you can type in Notepad in about fifteen minutes. Generators and plugins mostly scrape your existing pages and produce a link dump with no summary, which defeats the one genuine benefit of the exercise. If your platform makes uploading a root file awkward, a plugin solves that upload problem, not the writing problem.

How much should I pay someone to add llms.txt?

Very little. It is one short document and one file upload, so it should be absorbed into other website work rather than priced as a project. Treat a separate line item running into hundreds of pounds, and any monthly fee for maintaining it, as a sign you are being sold the letters AI rather than a piece of work. Nobody can promise a result from it, because no provider has confirmed reading it.

How do I check whether my llms.txt is working?

Type your address followed by /llms.txt into a browser and check it loads as plain readable text rather than downloading or showing a 404. Beyond that, the only real test is your server access logs: search them for llms.txt and see which user agents requested it. Ahrefs found the most frequent visitors were SEO audit tools at 21.7% of requests, and that Slackbot fetched llms.txt more often than PerplexityBot did.

Sources, all checked 15 September 2026. The llms.txt proposal and specification: llmstxt.org, proposed by Jeremy Howard of Answer.AI on 3 September 2024, version 2 published 10 August 2026 (required H1, optional blockquote, optional detail, H2 file lists; root or subpath placement). Google Search Central, "AI features and your website", last updated 10 December 2025, for the statement that no machine-readable files, AI text files or markup are needed. John Mueller on Bluesky, 17 June 2025, reported by Search Engine Roundtable, for "FWIW no AI system currently uses llms.txt". Gary Illyes on Bluesky, reported by Search Engine Roundtable on 1 August 2025, for the keywords meta tag parallel. Search Engine Land, 20 May 2026, for Google adding an llms.txt check to Chrome Lighthouse's Agentic Browsing category and for Google's statement that these audits do not concern Search rankings. Ahrefs, "We analysed 137K sites", published 15 June 2026, for 137,210 domains, 28% publishing a file, 97% receiving zero requests in May 2026, the 19.5% / 21.7% / 1.1% request breakdown, the Slackbot comparison and the finding that AI bots never request a file that does not exist. OpenAI's "Overview of OpenAI crawlers" developer documentation for GPTBot, OAI-SearchBot, ChatGPT-User and OAI-AdsBot. Anthropic's support documentation for ClaudeBot, Claude-User and Claude-SearchBot. Perplexity's bots documentation for PerplexityBot and Perplexity-User. Google Search Central's crawler documentation for Google-Extended and the statement that it is not a ranking signal. RFC 9309 (2022) for the formalisation of robots.txt. No public statement by Microsoft on whether Bing or Copilot reads llms.txt could be found; that is recorded on the page as unestablished rather than guessed at. No figure is published by any provider for how often llms.txt changes an AI answer, and none is estimated here.