llms.txt is a plain Markdown file at yoursite.com/llms.txt that gives AI agents a short summary of your site and links to its most useful pages. It was proposed by Jeremy Howard on 3 September 2024 and updated to version 2 in August 2026. It is a proposal, not an official standard.
Who reads it: coding agents and documentation tools use it most. Google says Google Search does not use it, and as of 22 October 2026 OpenAI and Anthropic do not say their assistants read it when answering. Make one if you like (our free llms.txt generator builds it in minutes), but it will not get your business recommended on its own.
What is llms.txt, in plain English?
Web pages are built for people. They wrap the useful text in menus, pop-ups, ads and scripts. An AI agent that wants the facts has to strip all that away, and its working memory (the "context window") is too small to read a whole site. The llms.txt proposal says: put one short file at a known address that tells the agent what the site is and which pages to read.
It sits next to two files you may already know:
| File | What it does | Who it is for |
|---|---|---|
| robots.txt | Says which crawlers may access which parts of the site. An official standard (RFC 9309) | Crawlers, before they fetch anything |
| sitemap.xml | Lists every page you want indexed | Search engines |
| llms.txt | A curated summary with links to the pages that matter, in Markdown | AI agents, "on demand, when an agent needs information about a topic while assisting a user" |
The proposal says llms.txt is "designed to coexist" with these files and that robots.txt and llms.txt "have different purposes". It cannot allow or block a crawler. If robots.txt or your firewall blocks an AI bot, an llms.txt file changes nothing.
The format, as the v2 proposal defines it
The file lives at the root path, /llms.txt, or at a subpath such as /docs/llms.txt, where it covers only the pages under that path. It contains these parts, in this order:
- An H1 with the name of the site or project. "This is the only required section."
- A blockquote (a line starting with
>) with a short summary containing the key facts needed to understand the rest. - Optional details: paragraphs or lists, but no headings.
- Zero or more H2 sections, each a "file list": a Markdown list of links, written
[name](url), optionally followed by a colon and a note. - An optional section called "Optional" for secondary links an agent can skip. Since v2 this is only a convention with no special meaning to tools.
Version 2 added three things. Pages can offer a clean Markdown copy at the same address with .md added (page.html.md) or swapped in (page.md). Pages can point to those copies with standard link tags: rel="alternate" type="text/markdown" for the Markdown version and rel="describedby" for the llms.txt file that covers the page. And a file now formally covers the pages under its path, with the most specific file winning.
A full example for a business
Most examples online are for software documentation. Here is one for a local business. The clinic, address and prices are made up; swap in your own facts, word for word as they appear on your site.
# Northside Family Dental
> Family and emergency dental clinic in Leeds, UK. Open 7 days, with same-day emergency appointments. We see private patients and children. This file lists our key pages for AI assistants.
- Address: 12 Example Street, Leeds LS1 0AA, United Kingdom
- Phone: +44 113 000 0000 (calls and WhatsApp)
- Hours: Monday to Saturday 8:00 to 20:00, Sunday 10:00 to 16:00
- Prices: check-up GBP 55, emergency visit GBP 95, hygienist GBP 70
- Languages spoken: English, Polish, Urdu
## Services
- [Emergency dentist in Leeds](https://example.com/emergency-dentist.md): same-day slots, what counts as an emergency, prices
- [Children's dentistry](https://example.com/children.md): ages, first visit, how we handle nervous children
- [Teeth whitening](https://example.com/whitening.md): options, prices, who it is not suitable for
## About the clinic
- [Our dentists](https://example.com/team.md): names, registration numbers, experience
- [Prices](https://example.com/prices.md): the full price list, updated October 2026
- [Find us](https://example.com/contact.md): map, parking, bus routes, accessibility
## Optional
- [Blog](https://example.com/blog.md): guides on dental care
Notice what it does: the facts a buyer asks about (where, when, how much, who for) sit in the summary and the first list, and every link has a note that says what is behind it. That is the part agents use. Our own file is live at adjoltz.com/llms.txt if you want a longer real one.
Skip the typing. Our free llms.txt generator builds this file from your site's name, summary and key pages, in the v2 format.
Who actually reads llms.txt?
This is where most articles overclaim. Here is what each company says, checked on 22 October 2026:
| Who | What they say or do | What it means for you |
|---|---|---|
| Google Search, AI Overviews, AI Mode | "Google Search itself doesn't use them." Creating one "will neither harm nor help" your visibility. It is "completely fine" to make one for other services | No effect on Google |
| OpenAI (ChatGPT) | Its crawler documentation explains robots.txt rules for GPTBot, OAI-SearchBot and ChatGPT-User and says nothing about reading site owners' llms.txt files. It publishes an llms.txt for its own developer docs | No documented use when ChatGPT answers buyers |
| Anthropic (Claude) | Its crawler page explains robots.txt rules for ClaudeBot, Claude-User and Claude-SearchBot and does not mention llms.txt. It publishes an llms.txt for its developer docs | No documented use when Claude answers buyers |
| Perplexity | Its crawler documentation for PerplexityBot and Perplexity-User does not mention llms.txt. It publishes one for its API docs | No documented use when Perplexity answers |
| Microsoft (Bing, Copilot) | Bing's Webmaster Guidelines, which cover Copilot, do not mention llms.txt | No documented use |
| Google Chrome Lighthouse | Its agentic browsing audit checks for the file. A server error fails the check; a missing file is "Not Applicable", because "providing the file is optional at the moment" | A quality check, not a ranking signal. Don't let the address return a server error |
| Coding agents and docs tools | The proposal says files "are used most heavily for software documentation, where coding agents follow them" | Matters if you sell software or an API |
There is a difference between publishing a file and reading other people's. The AI labs publish llms.txt so that developers' coding agents can find their API docs. None of them says its consumer assistant looks for your llms.txt before it recommends a plumber, a clinic or an agency.
Google's John Mueller put it bluntly in April 2025, as reported by Search Engine Journal: "AFAIK none of the AI services have said they're using LLMs.TXT", and he compared it to the old keywords meta tag. Eighteen months later, the documentation still says the same.
Can an assistant read it anyway? Yes, like any public page. If a user pastes your site into Claude or ChatGPT, the assistant's fetcher (Claude-User or ChatGPT-User) can open any URL it chooses, and a file at a known address is easy to find. That is a possibility, not something any of them documents.
What server logs show
Several companies have checked whether AI bots actually request llms.txt. None of these studies comes from an AI company, and two come from SEO tool vendors, so read them as signs, not proof:
| Study | What they measured | What they found |
|---|---|---|
| Flavio Longato, personal blog (Aug 2025) | 30 days of CDN logs for 1,000 Adobe Experience Manager domains | "No GPTBot, ClaudeBot, PerplexityBot, or similar were seen at all" requesting the file |
| OtterlyAI, vendor (Feb 2026) | 90 days of AI bot traffic on one test site | 84 of 62,100+ AI bot visits went to /llms.txt, about 0.1% |
| SE Ranking, vendor (2026) | Nearly 300,000 domains | 10.13% had the file; "no correlation between AI citations and LLMs.txt" |
The pattern is consistent: few sites have the file, few AI bots ask for it, and sites with it are not cited more. That matches what the AI companies document.
Common mistakes
- Treating it as a ranking factor. It is not one for Google, by Google's own statement, and no assistant documents using it for recommendations. If an agency sells you llms.txt as "AI SEO", ask what else they do.
- Using it instead of robots.txt. It cannot allow or block bots. Check robots.txt and your firewall first, because a blocked crawler never reaches either file.
- The wrong address or type. It must be at
/llms.txt(or a subfolder path), return status 200 and serve text, not your HTML 404 page. Open the address in a browser and check. - No H1, or HTML instead of Markdown. The H1 with your name is the one required part. Plain Markdown, no page layout.
- Facts that don't match the site. A price or address in llms.txt that differs from your pages teaches an agent two versions of you. Copy the facts word for word, and update both together.
- Links to pages agents cannot open. Pages behind a login, blocked in robots.txt or behind a bot challenge are dead ends.
- Dumping the whole site into it. The proposal says "the file itself stays small enough to fit in context" and the detail lives behind the links. Put long text on the linked pages.
- Writing for robots, not readers. Keyword lists and claims like "the best clinic in the city" add nothing an agent can use. Plain facts do.
Should you make one?
Worth doing if
- You sell software, an API or anything with documentation.
- You want one clean summary of your facts to point agents and partners to.
- You can keep it updated when prices or hours change.
Don't expect
- Better Google rankings or more AI Overview links.
- ChatGPT, Gemini or Claude to recommend you because of it.
- It to fix a site that blocks AI crawlers or hides prices in images.
Our position: we add a correct llms.txt for clients because it takes an hour and does no harm, and we tell them it is housekeeping, not the work that gets them named. That work is crawler access, one clear page per buyer question, and being listed on the sites AI cites. Our AI crawlers guide covers the access part.
What we'd do for you
Start with the free AI check: 10 of your buyers' real questions on ChatGPT and Gemini (Google AI), with the businesses named instead of you. The AI visibility audit (AED 1,950, or USD 549 for clients outside the GCC at adjoltz.com/us) checks crawler access, robots.txt, firewall rules and llms.txt as part of a full review across four AI tools. Monthly plans (Foundation AED 4,500 or USD 1,249, Growth AED 8,500 or USD 2,349, Authority AED 16,000 or USD 4,399) run month to month with no setup fee and an unconditional 30-day money-back guarantee (rules). We work with businesses worldwide.
What we don't promise
- That llms.txt will change what any AI says about you. Nobody can honestly promise that today.
- That the proposal will become a standard, or stay in its current form. It changed in August 2026 and may change again.
- That any tool "checks your llms.txt score" in a way AI systems care about. There is no such score.