What is llms.txt and does your site need one? The 2026 data

The llms.txt file is one of the most searched terms in SEO for AI. When it was proposed in 2024, nobody knew whether it would work. Now we do: two large studies measured it across hundreds of thousands of domains and Google said outright what it thinks of it. We explain what it is, what it looks like and how to add it. Then, straight out, what the measurements found, because that changes the answer to whether you need it.

What llms.txt is in one sentence

It is a plain text file at yoursite.com/llms.txt, which sums up in a simple format who you are, what you do and where on your site AI tools can find the most important things.

The name comes from the abbreviation LLM: large language model. That is the technical name for the technology behind ChatGPT, Gemini, Claude and other AI assistants.

A comparison with files you may already have on your site helps:

• robots.txt is the doorman: it tells bots where they may go and where they may not,
• sitemap.xml is the list of all the rooms: the addresses of every page on the site,
• llms.txt is a business card with the table of contents of a book: it says briefly what the site is about and what to read first.

The file was proposed in 2024 by Jeremy Howard of Answer.AI. It is a proposal, not an official standard: nobody requires it and nobody guarantees that anything will read it. Unlike robots.txt it forbids nothing and permits nothing, it is only a summary.

Why it came about

An ordinary web page is easy for a person to follow, but for AI it is full of clutter: menus, footers, cookie bars, scripts, ads. At the same time, an AI tool can only process a limited amount of text at once. When it looks for an answer, it needs to get to the point fast.

The llms.txt file serves it up on a plate: plain text with no design, with a short summary and links to the pages that really matter. It is written in Markdown. That is ordinary text in which a hash # marks a heading and a link is written as [name](url).

The idea is a good one. The question is whether anybody uses it.

What llms.txt looks like: an example

This is what the file could look like for a made up hair salon:

# Kaderníctvo Lúka

> A family hair salon in the Old Town district of Bratislava. Women's and men's
> cuts, coloring and wedding hairstyles. Book online or by phone.

Key facts:
- Address: Obchodná 1, 811 06 Bratislava
- Open: Monday to Friday 9:00 to 18:00, Saturday 8:00 to 12:00
- Phone: +421 900 000 000

## Services and pricing

- [Pricing](https://www.example.com/pricing): current prices for all services
- [Online booking](https://www.example.com/booking): available times

## About us

- [Our team](https://www.example.com/team): our hairdressers and what they specialize in
- [Contact](https://www.example.com/contact): address, parking, map

## Optional

- [Blog](https://www.example.com/blog): hair care tips

What the individual parts mean:

• Heading with a hash (# Meadow Hair Salon): the name of your company or website. In the proposal this is the only required part.
• Lines starting with >: a short summary in two or three sentences. The most important part of the whole file.
• Key facts: free text with the things AI should know for certain.
• Sections with two hashes (## Services and pricing): lists of links to important pages with a short description.
• The Optional section: less important links that AI can leave out when it runs short of space.

You can see a live example on our own site: meaovis.com/llms.txt. Our file is generated automatically from the same data as the website, so the prices and services in it are always current.

What has actually been measured about llms.txt

This is where the article turns. Most guides online still claim that llms.txt will raise your visibility in AI. When somebody wrote that in 2025 it was a guess, because there was no data. Now there is, and it says something else.

97 percent of the files were read by nobody

In June 2026 Ahrefs published a measurement covering 137 210 domains from its own analytics. The result: 97 percent of llms.txt files received not a single request in the whole of May 2026. All of the measured traffic went to the remaining three percent.

What is more interesting is who sent those requests. Of the visits that did reach the files, 96 percent came from bots, and the largest single group was SEO audit tools, at 21,7 percent. That is more than all AI bots put together, which came to 19,5 percent. AI retrieval bots, the very ones people are hoping will bring them visibility, accounted for 1,1 percenta.

Put another way: the tools that most often read an llms.txt file today are the tools checking whether you have one. Ours included, we admit it.

And one more figure that says more about the real level of interest than anything else. When the file is nie je on the site, requests to its non-existent address are 98 percent human and from AI bots there comes ani jedna. AI tools simply do not look for the file.

It has no effect on AI citations

SE Ranking examined the relationship between having the file and how often AI tools cite a domain. The sample was roughly 300 000 domains, of which 10,13 percent had the file. They found no significant association. The most telling detail: when llms.txt was removed from their prediction model, the model got more accurate. Their own conclusion is cautious rather than dismissive: it does not appear to affect citation frequency directly, at least not yet.

Google has said so openly

This is not guesswork, Google has its own guide to AI features in Search. It states that to appear in them you do not need to create any new machine readable files or AI text files, and that there are no additional technical requirements either. When John Mueller of Google was asked directly in January 2026 whether llms.txt helps, he answered no, and added that he was being direct about it.

In May 2026 Chrome added an llms.txt check to Lighthouse, which part of the internet read as backing from Google. It is not: the check sits in the category of browsing for AI agents, not search. In the Ahrefs measurement Lighthouse accounted for about 22 requests, roughly one in a thousand.

Where it does make sense

The same data brings out a second point you will read about less often. If you look at the named AI bots in the Ahrefs measurement, GPTBot comes first at 4,51 percent and na druhom Claude-Code, a developer tool rather than a search engine. It came ahead of every AI assistant and every AI search engine.

That says quite precisely what llms.txt is good for: it is documentation for AI that works with the site directly. If you have extensive product documentation, an API or a library and you want a programmer's AI assistant to find its way around it, llms.txt is useful and worth doing properly.

If you are a hair salon, a joinery or a law firm, none of that applies to you. For you it is a file that takes an hour, costs nothing and in all likelihood nobody will ever open.

How to add llms.txt to your website

If you still want one after all that, here is how:

1. Write down the key facts. Who you are, what you do, for whom, where, guide prices and contact details. Write plainly, without advertising phrases.
2. Pick your 5 to 15 most important pages. Services, pricing, contact, FAQ, references. Not every page on the site.
3. Write the file following the example in a plain text editor and save it in UTF-8 encoding so that accented characters display correctly.
4. Upload it to the root of your website so that it is available at yoursite.com/llms.txt. Your webmaster will do it over FTP or in the admin panel. For some content management systems there are extensions that create the file automatically.
5. Check it in a browser. Open the address of the file and check that it shows up and that the accented characters are correct.
6. Update it together with your website. Have you changed prices or services? Change llms.txt too. Best of all is when it is generated automatically.

Some websites also have a longer version called llms-full.txt with the full text of the most important pages. That is exactly what suits companies with extensive documentation. For an ordinary business even the short version is pointless, let alone the long one.

What llms.txt will not do

The llms.txt file is not a magic switch that “turns you on” in ChatGPT. Specifically:

• it will not guarantee that AI starts mentioning or recommending you, and according to a measurement across 300 000 domains it does not do so even on average,
• it does not replace good content, clear answers to questions or reviews,
• it does not help if your robots.txt blocks AI bots (they never reach the file at all),
• it will not improve your Google rankings, which Google has confirmed itself,
• it will not replace structured data, which Google and AI genuinely do read.

The most common mistakes

• Wrong address. The file sits in a subfolder or is accidentally named llms.txt.txt.
• Broken accented characters. The file is not saved in UTF-8 encoding.
• Outdated data. Services you no longer offer, old prices, a phone number different from the one on the site.
• Broken links to pages that no longer exist.
• The whole website in one file. The short version is meant to be short.
• Advertising slogans instead of facts. AI does not quote “the best services in town”.
• Addresses that are not meant to be public. The file is a list of links to your key pages, so do not put anything in it that is not in your sitemap: test versions, internal dashboards, a campaign in preparation.

So does your site need one?

Straight out, so you do not have to read between the lines:

• As a visibility tool, no. Two independent measurements and Google's own statements say the same thing. If somebody is selling it to you as the way into ChatGPT answers, they are selling you something they cannot back up.
• Paying for one, definitely not. The hour it takes to write one can be spent on three clear answers on your site instead, and those will work.
• If you are getting one for free, why not. It does no harm, and if it turns out in two years that AI agents really do read it, you will already have one.

That is why we generate one automatically for the sites we build and include it in the price. Not as an argument about visibility, but as an item that costs us nothing and costs the client nothing either. And it is why, in our Audite zdarma we list it for information only and do not count it towards the score: taking points off you for something that will bring you nothing would not be fair.

What to do instead

If you want AI tools to mention you, this is what works: let AI bots in, answer the questions customers actually type specifically on your site, get your structured data right, and collect reviews and mentions on other sites. None of that is a file; all of it is work on content.

Our Free audit goes through the technical side: whether AI bots have access, whether a firewall is blocking them, whether the page carries structured data and whether the site works at all without JavaScript. If you want certainty going forward, Website Guardian checks it automatically every week.

Zdroje

• Ahrefs: We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read, June 2026.
• SE Ranking: an analysis of the relationship between llms.txt and AI citations across roughly 300 000 domains.
• Google Search Central: the guide to AI features in Search (AI features and your website).
• A statement by John Mueller of Google, January 2026.