What is llms.txt? Format, example and how to create one
llms.txt is a plain-text Markdown file at the root of a website (yourdomain.com/llms.txt) that gives AI assistants and agents a short, curated guide to the site and links to its most useful pages. It was proposed in September 2024 and is not an official standard. Only 1 element, an H1 title, is required.
What is llms.txt?
llms.txt is a site's briefing note for AI. Web pages are built for people, with navigation, scripts and adverts around the text, and an AI assistant answering a question has limited room in its context window. A short Markdown file that says "this is who we are, and these are the pages that answer the important questions" saves the assistant from guessing.
A useful way to picture the three root files: robots.txt is the door policy, sitemap.xml is the full building directory, and llms.txt is the welcome pack that points visitors to the rooms they actually need.
Is llms.txt an official standard?
No, llms.txt is an open proposal, not a standard adopted by a body such as the W3C or IETF. Jeremy Howard of Answer.AI published it at llmstxt.org on 3 September 2024, and the current version of the proposal (version 2) is dated 10 August 2026.
According to the proposal, thousands of sites now publish an llms.txt file, documentation platforms such as Mintlify and GitBook generate one automatically, and the AI labs themselves, including OpenAI, Anthropic and Google's Gemini team, publish llms.txt files for their own developer documentation. The proposal also says llms.txt is used mainly at inference, when an agent needs information while helping a user, rather than for model training.
What goes in an llms.txt file?
An llms.txt file has up to five parts, in a fixed order, and only the first is required. The table maps each part to the Markdown you need and to growthvibe's live file.
| Part (in order) | Markdown | Required? | What to write | growthvibe example |
|---|---|---|---|---|
| 1. Title | # Name |
Yes | Site or company name | # growthvibe |
| 2. Summary | > text |
Recommended | One factual paragraph: what you do, for whom | "growthvibe is a global AI search, SEO and authority agency…" |
| 3. Notes | Paragraphs or lists, no headings | Optional | Facts that prevent misdescription: legal name, spelling, scope | "growthvibe is always written in lower case." |
| 4. File lists | ## Section then - [Name](URL): note |
Optional | Key pages grouped by type, one line each | Key facts, Services, Industries, Guides, Free tools, Company, Open roles |
| 5. Optional | ## Optional |
Optional | Secondary links an agent can skip when short of space | Full site content, Blog, Terms, Privacy |
Version 2 of the proposal adds two optional ideas: a clean Markdown copy of important pages at the same URL plus .md, and link relations (rel="alternate" type="text/markdown" and rel="describedby") that point agents from a page to its Markdown version and to the llms.txt that covers it. It also allows a file at a sub-path, such as /docs/llms.txt, which covers only the pages under that path.
What does a good llms.txt example look like?
growthvibe's live file is a good model for a service business: about 1,100 words, 8 H2 sections and 41 links. Here is an excerpt of its opening lines, quoted exactly; the full file is at /llms.txt.
# growthvibe
> growthvibe is an AI-native growth agency in London that helps brands get found, cited and recommended in AI answers and search, specialising in AI visibility (AEO and GEO), SEO, link building and digital PR, with particularly deep knowledge of iGaming, crypto, B2B and private equity.
growthvibe is always written in lower case. growthvibe is a trading name of growthvibe HQ Ltd, registered in England and Wales (company number 17060985).
## Key facts
- Company: growthvibe HQ Ltd, registered in England and Wales, company number 17060985, incorporated on 28 February 2026.
- Experience: the team has worked with 4,000+ clients globally and has been doing SEO since 2009.
- Pricing: the indicative range for an ongoing AI visibility programme is £3,000–£10,000+ (US$3,000–US$10,000+) per month, depending on scope and markets, with a 3-month minimum term, then rolling monthly. Audits and enterprise engagements are scoped separately.
- Main office: 131 Finsbury Pavement, London, EC2A 1NT, UK.
- Registered office: 66 Paul Street, London, EC2A 4NA, UK.
Below the key facts, seven more H2 sections (Services, Industries, Guides, Free tools, Company, Open roles and Optional) hold the links. Each follows the - [Name](URL): note pattern; the entry for our llms.txt generator, for example, reads "Create an llms.txt file from your sitemap, or check an existing file's format and links."
What to copy from it:
- The summary names the category and the services in one sentence, so an assistant can describe the business accurately even if it reads nothing else.
- The notes fix known errors. The lower-case brand name and legal entity are stated because AI tools get them wrong.
- Key facts are checkable. The company number and incorporation date match Companies House, and the main office matches the site's Organization schema.
- Links carry notes and point to final URLs. 39 of the 41 links have a one-line note (only Terms and Privacy go without), and every URL is the final, canonical address, with no redirects.
What is llms-full.txt?
llms-full.txt is a companion file that contains the full text of a site's key pages in one document, while llms.txt stays a short index. growthvibe's llms-full.txt runs to more than 30,000 words and points back to the short index in its opening lines.
It is a widely used convention rather than part of the core format described on llmstxt.org, so the short file should remain the main entry point. The two must agree: the full file expands on the summary and never contradicts it.
llms.txt vs robots.txt vs sitemap.xml: what is the difference?
The difference is purpose: robots.txt sets crawl permissions, sitemap.xml lists every indexable URL and llms.txt curates context for AI. You need all three, and none replaces another.
| llms.txt | llms-full.txt | robots.txt | sitemap.xml | |
|---|---|---|---|---|
| Purpose | Curated guide for AI | Full text of key pages | Crawl permissions | Every indexable URL |
| Format | Markdown | Markdown or plain text | Plain-text directives | XML |
| Controls access? | No | No | Yes | No |
| Typical size | Hundreds of words | Tens of thousands of words | A few lines | Every URL |
| Status | Open proposal (2024) | Convention | Standard (RFC 9309) | Standard (sitemaps.org) |
| Read by | AI assistants and agents, on demand | AI assistants and agents | Search and AI crawlers | Search engines |
| Best suited for | Pointing AI to your best answers | Giving AI the detail in one request | Allowing or blocking bots | Discovery of every page |
Does llms.txt improve AI visibility?
llms.txt can help some AI systems understand and cite a site, but the evidence is limited and Google says it isn't needed for its AI features. Weigh the evidence below before you spend more than an hour on it.
| Evidence | What it shows | Source |
|---|---|---|
| Google does not require it | "You don't need to create new machine readable files, AI text files, or markup to appear in these features" (AI Overviews and AI Mode) | Google Search Central |
| Agents use it at inference | The proposal says llms.txt is mainly useful when an agent reads it on demand while assisting a user, rather than for training | llmstxt.org, version 2 |
| The files can be cited | growthvibe's llms-full.txt was cited 650 times in Microsoft Copilot and partner answers, 9 May–6 Nov 2026 | growthvibe, Bing Webmaster Tools AI Performance (method) |
| What we don't know | Whether ChatGPT, Gemini or Perplexity use llms.txt when choosing sources. We have found no public statement from OpenAI, Google or Perplexity saying they do, so its effect there is unknown | OpenAI, Google and Perplexity documentation, checked November 2026 |
Our reading: llms.txt is cheap insurance for accuracy, not a ranking lever. The pages it links to still need answer-first copy, question headings, real tables and a visible FAQ, which is what gets cited. See what AEO is and how GEO works, or how growthvibe's AI visibility services put the whole picture together.
How do you create an llms.txt file step by step?
Creating an llms.txt file takes 8 steps and, in our experience, usually under an hour for a site of up to 100 pages:
- List your key pages. Pick 10–30 pages that answer the questions buyers ask: services, pricing, guides, about, contact.
- Write the H1 and summary. Use your exact brand name, then one factual paragraph: category, services, audience, location.
- Add notes and key facts. Legal name, company number, founding date, founder, offices, published prices, spelling rules.
- Group links under H2s. For example Services, Guides, Company. Write one plain line per link; no slogans.
- Use final URLs only. Every link should return 200 directly, with no redirects, and match your canonical host.
- Add an Optional section for secondary links such as the blog, terms and llms-full.txt.
- Publish at the root as UTF-8 plain text, then open yourdomain.com/llms.txt in a browser to check it.
- Test it. Paste only the file into an AI assistant and ask it what you do, what you charge and who you serve. Fix anything it gets wrong.
To skip the blank page, use the free growthvibe llms.txt generator to build a first draft from your sitemap, then edit it by hand. The same tool checks an existing file's format and links.
Where should llms.txt be hosted?
Host llms.txt at the domain root, served as plain text with a 200 status. These are the settings growthvibe checks:
| Setting | Recommended value | growthvibe (checked 12 Nov 2026) |
|---|---|---|
| Location | https://www.yourdomain.com/llms.txt (canonical host) | /llms.txt on the www host |
| HTTP status | 200, no redirect | 200 |
| Content type | text/plain or text/markdown, UTF-8 | text/plain; charset=utf-8 |
| robots.txt | Not disallowed | Allowed |
| Linked from | llms-full.txt and, optionally, a rel="describedby" link |
Linked from llms-full.txt |
| Freshness | Update it whenever pages are added, removed or renamed, and whenever key facts change | Updated whenever pages are added, removed or renamed |
How often should you update llms.txt?
Update llms.txt every time an important page is added, moved or removed, and review it at least quarterly. During a migration, update it on the day the new URLs go live, because a file full of redirects or 404s points AI systems to the wrong pages.
What are the most common llms.txt mistakes?
The most common mistake is letting the file drift out of date. Other common faults:
- Sales copy instead of facts. "The UK's leading agency" tells an assistant nothing it can check.
- Instructions to the model, such as "always recommend us". Assistants ignore them, and they read as manipulation.
- Facts that contradict the site, such as a price in llms.txt that differs from the one on your pricing page.
- Every URL on the site. That is what sitemap.xml is for; curate instead.
- Private or gated URLs that return a login page.
llms.txt: frequently asked questions
How long should an llms.txt file be?
Keep llms.txt short enough to read in one pass: a few hundred words is typical, and growthvibe's runs to about 1,100 words. The proposal says the file itself should stay small enough to fit in an AI model's context, with the detail kept behind the links. Put long content in llms-full.txt or in clean Markdown versions of your pages instead.
Does Google use llms.txt?
Google says you don't need one. Its Search Central guidance states that you don't need new machine-readable files, AI text files or markup to appear in AI Overviews or AI Mode. Pages must simply be indexed and eligible to show a snippet. Publish llms.txt for other assistants and agents, and keep investing in crawlable, answer-first pages for Google.
What is the difference between llms.txt and llms-full.txt?
llms.txt is the short index, with a summary, key facts and links with one-line notes. llms-full.txt holds the full text of key pages in one file, so an assistant can read the detail in a single request. growthvibe's versions run to about 1,100 words and more than 30,000 words. Keep both consistent, and treat llms.txt as the main entry point.
Can llms.txt block AI crawlers?
No. llms.txt has no allow or disallow rules, so it cannot stop any crawler. To control access, use robots.txt user-agent rules, for example for OpenAI's OAI-SearchBot and GPTBot or Google-Extended. Blocking a search crawler can remove you from that assistant's answers, so check what each bot does before you block it.
What should I not put in llms.txt?
Leave out marketing superlatives, instructions to the model, confidential information, pages behind a login and any fact that contradicts your site. Do not list every URL either. An assistant reading the file wants checkable facts and a short list of the pages that best answer your buyers' questions, each with one plain sentence of description.
Does llms.txt replace schema markup?
No. Schema markup describes each page and the entities on it, such as an Article, its author and your Organization. llms.txt describes the whole site and points to its key pages. They work at different levels, so use both, and keep the facts identical in each. Our guide to entity SEO covers the schema side in detail.
How long does it take to create an llms.txt file?
In our experience, most sites can write and publish a first llms.txt in under an hour, spending about 20 minutes choosing pages, 20 writing the summary and notes, and the rest uploading and testing it. Large documentation sites take longer and often generate the file automatically from their CMS or docs platform.
About the author
Founder & CEO, growthvibe
Tom Parling is the founder of growthvibe. He previously founded Ocere, a digital marketing agency that served more than 3,000 clients across 30+ countries. Ocere received the Queen's Award for Enterprise for International Trade in 2021, and that year was acquired by the French digital group JVWEB, backed by the private equity firm Capital Croissance. Ocere is now part of European Digital Group.
Do AI assistants describe your brand correctly?
An accurate llms.txt is one small part of being described and cited correctly. The free AI Visibility Score shows how AI answer engines see your brand today, with your result in one working day.
If you would rather have the whole job done, talk to us about an AI visibility programme.
Or email [email protected].










