llms.txt: What It Is, an Example, and Does It Work
llms.txt is a proposed standard for a plain Markdown file, placed at /llms.txt on a website, that gives AI tools a short summary of the site and links to its most useful pages. It's cheap to add and does no harm, but as of November 2026 we haven't found any AI assistant that has publicly confirmed it uses llms.txt when deciding what to say or which brands to recommend.
This post explains what llms.txt is, shows an example for a B2B supplier, and sets out what's known about whether it works.
What is llms.txt?
The llms.txt file was proposed by Jeremy Howard of Answer.AI in September 2024, with the specification at llmstxt.org. A second version of the proposal was published there in August 2026.
The idea is simple. Web pages are built for people, wrapped in navigation, scripts and layout, and an AI agent reading a site has limited room in its context window. So the site owner writes one short file that says what the site is and points to clean, text-only versions of the pages that matter.
A few things it is not:
- It isn't robots.txt. robots.txt tells crawlers what they may access. llms.txt doesn't grant or block anything; it's a guide for tools that have already chosen to read your site. The llmstxt.org page makes the same distinction.
- It isn't a sitemap. A sitemap lists every page. llms.txt is a curated shortlist, with notes.
- It isn't a ratified standard. It's a proposal maintained in a public GitHub repository, open to community input.
The llms.txt format
According to the specification, the file is Markdown with sections in a fixed order:
- An H1 with the name of the site or project. This is the only required part.
- A blockquote with a short summary containing the key facts needed to understand the rest.
- Optional paragraphs or lists with more detail, but no headings.
- Optional H2 sections, each holding a list of links in the form
[name](url), optionally followed by a colon and a note.
A section headed "Optional" is, by convention, for links an agent can skip when it's short of space. Version 2 also proposes that pages offer a Markdown copy at the same address with .md added, and that pages link to their llms.txt file with standard link relations. The file can sit at the root of the site or in a subfolder such as /docs/llms.txt, covering the pages beneath it.
An llms.txt example for a B2B supplier
Here's an llms.txt example for an invented industrial distributor. The company and its addresses are made up for illustration.
# Example Fasteners Ltd
> Example Fasteners is a UK distributor of industrial fasteners and fixings
> for manufacturers and construction firms, with next-day delivery across
> mainland UK from warehouses in Leeds and Bristol.
We stock around 40,000 product lines, including stainless steel and
high-tensile bolts, anchors and rivets. Trade accounts only; no retail sales.
## Products
- [Stainless steel bolts](https://www.example.com/bolts/stainless.md): Grades A2 and A4, M3 to M36
- [Chemical anchors](https://www.example.com/anchors/chemical.md): Resins and studs for concrete and masonry
## Buying from us
- [Opening a trade account](https://www.example.com/trade-account.md): Credit terms and how to apply
- [Delivery and lead times](https://www.example.com/delivery.md): Cut-off times and regions served
- [Certifications](https://www.example.com/quality.md): ISO 9001 and material traceability
## Optional
- [Company history](https://www.example.com/about.md)
Notice what it does: it states the facts a buyer, or an assistant answering a buyer, would need first. Who you sell to, what, where, and on what terms. If you only take one idea from llms.txt, take that one, because those facts belong on your ordinary pages too.
Our free llms.txt generator builds a file in this format from your site's details, so you can see what yours would look like.
How to add llms.txt to your site
If you decide to publish one, it's a small job:
- Write the summary first. One or two sentences that say who you are, what you sell, who to, and where. If you can't write this crisply, fix that on your home page too.
- Pick ten to twenty links, not two hundred. Choose the pages a buyer would need to make a decision: product categories, delivery, terms, certifications, contact. The file is meant to be short enough to read in one go; the detail lives behind the links.
- Add a note to each link saying what's on the page, so a tool can decide whether to follow it without opening it.
- Publish it as plain text at
/llms.txton your main domain. Most website platforms let you upload a file to the root, and some, such as Wix and the Yoast SEO and AIOSEO plugins for WordPress, can generate one for you. - Keep it current. A file that points to old pages or discontinued products does more harm than having none. Put a reminder in the calendar to review it when your range or prices change.
- Test it. The specification suggests giving an AI agent only your llms.txt as a starting point and asking it questions about your business. If it can't answer the basics, the file needs work.
The Markdown copies of pages that version 2 proposes are optional. They're most useful for documentation sites; for a typical supplier's marketing site, clear HTML pages with the key facts in text are what matters.
Does llms.txt work?
It depends what you mean by "work".
What's known
- Some tools read it. The proposal says llms.txt is used most heavily for software documentation, where coding agents follow it to find reference pages. Documentation platforms and CMSs listed on llmstxt.org, including Mintlify, GitBook, Yoast SEO and Wix, can generate one.
- The AI companies publish them for their own developer docs. The llmstxt.org page lists OpenAI, Anthropic and Gemini. Publishing a file for developers' coding tools isn't the same as using other sites' files to answer questions, and none of those companies' crawler documentation that we checked mentions reading llms.txt.
- Google says you don't need it. Google's guidance on AI features and your website says you don't need to create new machine-readable files or AI text files to appear in AI Overviews or AI Mode.
- Lighthouse now checks for it. Chrome's Lighthouse has an llms.txt audit in its agentic browsing checks. A missing file is marked not applicable rather than failed, because the file is optional. That's a check that a file is well formed, not a sign that any assistant ranks you higher for having one.
What isn't known
We looked for a primary source, meaning OpenAI's, Anthropic's, Google's or Perplexity's own documentation or announcements, saying that their assistants use llms.txt when composing answers or choosing which brands to name. As of November 2026 we didn't find one. You'll see blog posts claiming otherwise; check whether they link to the company's own statement.
What we'd conclude
llms.txt is low cost and low risk. It may help AI agents and coding tools that read it, and it's a useful exercise in writing down your key facts. But adding one won't, on any evidence we've seen, get your brand recommended by ChatGPT, Claude, Gemini or Perplexity. If you have limited time, the work below matters more.
What matters more than llms.txt
- Let the right crawlers in. If the AI companies' search crawlers can't read your site, an llms.txt file won't help. Our guide to ClaudeBot, GPTBot and PerplexityBot explains which is which.
- Put your key facts in your pages as text, not only in images, PDFs or scripts.
- Be present on the sites assistants cite in your category: directories, marketplaces, trade bodies and comparison pages. In our published example report, the most cited site was a HubSpot marketplace listing, in 21 of 48 answers, while the brand's own site was cited once. The brand's site did have an llms.txt file; it didn't make the site the main source.
- Measure what assistants say. Our free AI visibility checker asks four assistants 12 buyer questions in your category and shows whether you're named and which sources they cite.
For the bigger picture, read how to get cited by ChatGPT. If you'd like us to work through it with you, see our AI visibility audit and retainer, or book a free call.