What is llms.txt, and does my business need one?
Short answerllms.txt is a plain-text file at the root of a website (yoursite.com/llms.txt) that summarises the site in Markdown and links to its most useful pages, so AI tools can understand it quickly. It is a proposal, not a standard. It is cheap and harmless to add, but no major AI search tool says it uses the file to decide which businesses to recommend.
Key points
- llms.txt is a proposed file, first published by Jeremy Howard in September 2024, that gives AI tools a short, curated summary of a website.
- It is used most by coding assistants and documentation tools; Google says you do not need it to appear in AI Overviews or AI Mode.
- It does not control crawling or AI training. That is still the job of robots.txt.
- Adding one takes under an hour and costs nothing, so it is good housekeeping, but it should never be sold to you as a strategy.
You have probably heard about llms.txt from someone offering to add one for you, or from a checklist that says every site now needs it. The truth is less exciting and more useful: it is a small, sensible file with a narrow job. This guide explains what that job is, who actually reads the file, and how to write one yourself without paying anyone.
What exactly is llms.txt?
llms.txt is a text file placed at the top level of a website, so it lives at an address like yoursite.com/llms.txt. The name comes from LLMs, large language models, the technology behind ChatGPT, Claude, Gemini and similar tools.
It was proposed by Jeremy Howard, an AI researcher, in September 2024, and the proposal at llmstxt.org was revised to a second version in August 2026. The idea is simple. Web pages are built for people, with menus, banners, scripts and pop-ups. An AI tool trying to understand a site has to dig through all of that. llms.txt offers it a clean summary in one place instead.
The file is written in Markdown, a plain-text format where a line starting with # is a heading and [text](link) is a link. Under the proposal, only one thing is required: a top heading with the name of the site. Everything else is recommended rather than mandatory.
The proposal also suggests offering a clean Markdown copy of each important page, at the same address with .md added, so tools can read the text without the design around it. That makes sense for large technical documentation sites. For a five-page business website it is optional extra work with no known payoff.
What does one look like?
This site has one, so here is how it is built, part by part:
| Part | What it contains on this site | Its job |
|---|---|---|
# M/AFZAL | The name of the practice | The one required line |
A quoted summary (a line starting with >) | Two sentences on what the practice does and who leads the work | The answer to "what is this site?" |
| A short paragraph | Where clients are based and how engagements run | Context the summary leaves out |
## Pages | Links to the home page, the AI search guide and the founder page, each with one line on what it covers | Points the tool to the pages that matter |
## Packages and ## Services | The three packages with exact prices, and the services offered | Facts a tool might be asked about |
## Contact | Website, email and LinkedIn | How to reach the business |
The whole file is under 500 words. That is the point: it is a briefing note, not a copy of the website.
Who actually reads llms.txt?
This is where most of the confusion lives, so it is worth being precise.
- Coding assistants and documentation tools use it. The proposal says it has been used mainly when an AI tool is answering a question, rather than for training models, and its heaviest users are software developers pointing coding assistants at technical documentation. Several documentation platforms now generate the file automatically.
- The AI companies publish their own. OpenAI, Anthropic and Google publish llms.txt files for their developer documentation. Publishing one is not the same as reading everyone else's to decide what to recommend.
- Google's Lighthouse checks for one. Lighthouse, the free site-testing tool built into Chrome, added llms.txt to a newer set of "agentic browsing" checks, which test how easily AI agents can use a site.
- Google Search says you do not need it. Google's guidance on AI features states that you do not need to create new machine-readable files, AI text files or markup to appear in AI Overviews or AI Mode.
- ChatGPT and Perplexity say nothing either way. Neither OpenAI's nor Perplexity's crawler documentation mentions llms.txt as something their search products use.
No major AI search tool documents using llms.txt to decide which businesses to name. Anyone who tells you it will get you recommended is guessing.
That could change. AI agents that browse, compare and book on a person's behalf are the use the file was designed for, and if they become common, a clear llms.txt may start to matter for ordinary businesses. For now, it is a bet that costs an hour.
How is it different from robots.txt and a sitemap?
Three files sit at the root of most websites, and they are often muddled together.
| File | What it does | Who it is for | Is it a standard? |
|---|---|---|---|
robots.txt | Says which crawlers may visit which pages | Every search engine and AI crawler | Yes, followed by all the major crawlers |
sitemap.xml | Lists every page you want found | Search engines | Yes, supported by Google and Bing |
llms.txt | Summarises the site and links its key pages | AI tools and agents | No, a proposal some tools follow |
The important consequence: llms.txt cannot block anything. If you want to stop AI companies using your content to train models, or to let their search crawlers in, that is done in robots.txt. The guide on whether to block AI crawlers covers which lines to use.
Should your business add one?
Yes, if it costs you an hour and nothing else. It is harmless, AI agents that browse on a buyer's behalf may use it, and it forces a useful discipline: describing your business in a few plain sentences. If you cannot write a clear llms.txt, your homepage probably is not clear either.
No, if someone wants to charge you a meaningful sum for it, or presents it as the centrepiece of an AI visibility plan. The work that is known to matter comes first: plain statements on your own pages, consistent facts across the web, structured data that labels those facts, and pages that answer buyer questions. The AI search visibility guide sets those out in order of value, and GEO vs SEO explains why most of it is ordinary search work done well.
As a guide to cost: writing the file is an hour of your own time, or a few minutes for whoever already maintains your website. If it appears on a quote as a separate line item with a meaningful price, ask what else on that quote has been padded the same way. And if your site runs on a builder that cannot host a file at the root, do not switch platforms for it; nothing known about llms.txt justifies that.
How do you write one?
- Start with the name. One line:
#followed by your business name exactly as it appears everywhere else. - Write the summary. One or two sentences on a line starting with
>: what you do, for whom, where, and one fact that proves you are credible. - Add the essential facts. A short paragraph with areas served, how the work runs, and anything buyers always ask.
- List your key pages. Under a
## Pagesheading, link each important page with one line saying what it answers. Use full addresses starting with https. - Add services and contact details. The same wording as your website, so the two never disagree.
- Upload it to the root of your site and check it opens at
yoursite.com/llms.txt.
Sample business: Hartwell Joinery. Its file would open with # Hartwell Joinery, then a summary line: "Hartwell Joinery designs, makes and fits bespoke kitchens in Leeds and Harrogate, in its own workshop since 2009, rated 4.9 from 128 reviews." Below that, a ## Pages section linking the prices page, the process page and the bespoke versus fitted comparison, then the workshop address and phone number exactly as they appear on its Google profile.
What mistakes should you avoid?
- Contradicting your website. An old phone number or a service you dropped in llms.txt adds to the inconsistency AI tools already punish.
- Pasting the whole site in. The value is in the curation. A wall of text defeats the purpose.
- Stuffing it with keywords. Repeating "best kitchen fitter Leeds" makes the summary less useful to a machine, not more.
- Publishing it while robots.txt blocks AI crawlers. A welcome note on a locked door helps nobody.
- Forgetting it exists. Put it on the same checklist as your Google Business Profile, so it changes when your details change.
llms.txt is housekeeping, not strategy. Do it once, do it accurately, then spend your effort on the work AI tools are known to reward.
Straight answers.
The follow-up questions owners ask most.
- Will llms.txt get my business recommended by ChatGPT?
- There is no public evidence that it will. OpenAI does not document reading llms.txt to choose what appears in ChatGPT answers, so treat it as tidy-up work rather than a ranking factor.
- Is llms.txt the same as robots.txt?
- No. robots.txt tells crawlers which pages they may visit. llms.txt describes what the site is and which pages matter most; it grants and blocks nothing.
- Does my website platform create llms.txt automatically?
- Some documentation platforms do, and some plugins for WordPress and other builders offer it. Check what any generated file says, because an automatic summary can be vague or out of date.
- How often should I update llms.txt?
- Whenever a fact in it changes: a new service, a new address, a renamed page. A file that contradicts your website does more harm than having none.