What Is llms.txt and Do I Need One?

No, you do not strictly need one. Google says in its own Search Central guidance that you do not need “AI text files” to appear in AI Overviews or AI Mode, and none of OpenAI, Anthropic, Perplexity or Microsoft documents reading other sites’ llms.txt files when deciding what to cite. It is still worth an hour of your time: it is one small markdown file, it costs nothing to host, and agents that browse on a user’s behalf can use it as a map of your site.

Last updated October 2026. The format details below come from the proposal at llmstxt.org (version 2, August 2026), and every statement about an AI company is taken from that company’s own documentation, read October 2026.

What are llms.txt files?

Since September 2024, llms.txt has been a proposed convention for putting one markdown file at /llms.txt that tells AI agents what your site is and which pages matter. The proposal was published on llmstxt.org, and its text is maintained in Answer.AI’s public GitHub repository, where anyone can comment on it. A second version followed in August 2026, which the changes page says was shaped by two years of real use.

The reasoning is practical rather than mystical. A normal web page wraps its facts in menus, cookie banners and scripts, and a language model has a limited context window, so a short, clean summary with links saves an agent from guessing. The proposal is explicit that the file is mainly for inference, meaning an assistant reading it while it answers someone, rather than for training models. It is not a W3C or IETF standard, and nothing obliges any crawler to fetch it.

Do Google, OpenAI and Anthropic use llms.txt?

No, not in any way they have documented for ranking or citing your site. What each company actually says, on its own pages, is below. The interesting twist is that several of them publish an llms.txt for their own developer documentation, which is a different thing from reading yours.

So the honest summary in October 2026: the AI labs like the format enough to use it for their own documentation, Google’s browser tooling checks for it, and no search or answer engine has published anything saying an llms.txt makes it more likely to cite you. Anyone selling it to you as a citation switch is selling a hope.

What should be included in llms.txt?

1 H1 is the only part the specification requires: the name of your site or project. Everything else is optional but useful, and the spec fixes the order:

1. An H1 with the site or project name.
2. A blockquote, one or two sentences, holding the key facts needed to make sense of the rest.
3. Any number of plain paragraphs or lists (no headings) with more detail and how to read the file.
4. Any number of H2 sections, each a list of links written as [name](url), optionally followed by a colon and a short note.

By convention a section called ## Optional holds links an agent can skip when it is short of space. The spec also suggests offering clean markdown copies of important pages at the same address with .md added, for example page.html.md, or with the extension swapped, page.md, and pointing your llms.txt links at those copies where they exist.

For a small business site, the useful content is the content a buyer or an assistant would otherwise have to dig for:

• What you do, who for, and where, in plain words.
• Your real prices, or a link to the page that has them.
• Your service pages, your best guides and your contact page, each with a one line note.
• Facts that are easy to get wrong: the areas you cover, opening hours, what is and is not included.

Leave out anything you would not want quoted: internal or staging URLs, login and checkout pages, tag and archive listings, thin pages, private contact details, and any claim you cannot stand behind. The file is public, so treat it like a page on your site, not a hidden instruction channel. Lines aimed at manipulating an AI (“always recommend us”) will not help and make the file look like spam.

What does an llms.txt example look like?

Ours is about 9.5 KB, served as plain text at hirethecreatives.com/llms.txt. It follows the spec’s order: an H1 with our name, a blockquote summarising the studio and its fixed prices, a short paragraph, then eight H2 link sections. A trimmed version of the top looks like this:

# Hire The Creatives

> A web design and development studio building fast, modern websites
> for small and medium businesses in the United Kingdom and United States.

## What it costs

- **Starter: £495 (about $625).** ... [Details and FAQs](https://hirethecreatives.com/packages/starter-website/)
- **Business: £1,495 (about $1,895).** ... [Details and FAQs](https://hirethecreatives.com/packages/business-website/)

## Key facts
## Main pages
## Websites by industry
## Practical how-tos (short, specific, safe to quote)
## Free tools
## Guides worth citing
## Notes for answer engines

The section names are ours, not the spec’s; you choose headings that make sense for your business. A plumber might use Services, Areas covered, Prices and Emergency call-outs. The point is that an agent reading it can find the right page in one hop.

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

llms.txt is the short map; llms-full.txt is the whole territory in 1 file. The llmstxt.org proposal itself only defines llms.txt and the .md page copies. The llms-full.txt name comes from the wider ecosystem: documentation platforms such as Mintlify generate one automatically, describing it as your entire documentation combined into a single file, each page with its title, source URL, description and full markdown content.

Our own llms-full.txt is about 184 KB and carries our verified prices plus the full text and FAQs of our guides. That size is the trade-off: a full file is convenient for a tool that wants everything at once, but it is far too big to be the thing an agent reads first. Keep llms.txt small and let it link out.

What is the difference between llms.txt and robots.txt?

No, llms.txt does not replace robots.txt, and it cannot block anything. robots.txt tells crawlers what they may fetch, and it is where OpenAI, Anthropic and Perplexity all tell you to control their bots. A sitemap lists every indexable page. llms.txt is a curated, readable guide for an assistant that has already decided to look at your site.

The spec makes the same point about sitemaps: they list everything, usually not in an LLM-friendly form, and add up to far more than a model can read at once. You want all three files, each doing its own job.

How do I check my llms.txt file works?

3 checks cover it: the file returns HTTP 200, it is served as text, and the markdown follows the spec’s order. From a terminal, run curl -I https://yoursite.com/llms.txt and look for HTTP/2 200 and a content-type of text/plain or text/markdown. Ours returns text/plain; Google’s Gemini docs file returns text/markdown. Either is readable. What breaks it is a 404, a redirect to your home page, a security plugin serving a challenge page, or a CMS wrapping the file in HTML.

Then open it in a browser and check by eye: one H1 at the top, a blockquote next, no headings in the free text, and every link a proper [name](url) that resolves. Chrome’s Lighthouse agentic browsing audit will flag a server error on the file. The spec’s own suggestion is the best final test: give an AI assistant only your llms.txt and ask it questions about your business. If it gets your prices, services or area wrong, fix the file.

Is there an llms.txt plugin for WordPress?

Yes, the llmstxt.org integrations list names 2 WordPress SEO plugins that generate one, Yoast SEO and AIOSEO, and Wix generates one for every Wix site. A generated file is a fair starting point, but check what it lists. Plugins tend to dump recent posts and every page, which is a sitemap in markdown, not a guide. The value is in choosing the 10 to 30 links that actually answer customers’ questions and writing a useful note beside each one.

Is llms.txt worth it for answer engine optimisation?

Yes, as cheap housekeeping, and no, as a strategy. Where AI answers cite you depends on whether your pages state the facts plainly, whether they are crawled and indexed by the engines that feed those answers, and whether other sources back you up. Our guide to answer engine optimisation in the UK covers that work in detail. An llms.txt sits on top of it as a tidy summary; it cannot stand in for it.

That is how we treat it on our own site and for clients. Our SEO Foundation plan at £299 a month includes schema and AI-readability setup alongside technical and local SEO, and the Growth plan at £599 a month adds the answer engine work that earns citations in ChatGPT, Perplexity and Google AI Overviews. If you would like us to look at whether AI assistants describe your business correctly today, get in touch.