AI
llms.txt: what the file does and how to create one
Since 2024, a new file has been turning up in SEO checklists: llms.txt. Some providers sell it as the key to visibility in ChatGPT, others consider it pointless. This guide covers what goes into the file, who actually requests it, what the latest data from Google, Ahrefs and SE Ranking shows and how to create your own. Our own llms.txt serves as the example.
What is an llms.txt?
An llms.txt is a text file in Markdown format located at yourdomain.com/llms.txt. In a few lines, it explains to AI systems who is behind the website and links to the most important pages, each with a short description. LLM stands for large language model, meaning language models such as GPT, Claude or Gemini.
The idea comes from Jeremy Howard, co-founder of the AI lab Answer.AI. He published it in September 2024 at llmstxt.org. The background: normal web pages consist largely of navigation, scripts and design. A language model can only process a limited amount of text per request and first has to filter out the actual content. The llms.txt provides a compact overview without this clutter.
The llms.txt is a proposal, not an official standard. robots.txt, by contrast, has been an established internet standard since 2022 (RFC 9309). Since August 2026, the llms.txt proposal has been at version 2.
How an llms.txt is structured
The proposal sets a fixed order. Only the first line is mandatory.
- Heading with the name of the website or company, as a first-level Markdown heading (
#). - Summary as a blockquote (line starts with
>): one or two sentences on what you do and for whom. - Further details as normal text, for example address, year founded and contact details. Headings are not intended here.
- Sections with link lists, each with a subheading (
##). Every entry follows the pattern- [Name](URL): Description. - An “Optional” section for secondary content. In version 1 it was meant for an AI system to drop when space is short. In version 2 it is still allowed but no longer has a special function.
Here is our own llms.txt, heavily shortened and translated into English:
# Media on Earth
> Websites, web apps and SEO for businesses that want to win enquiries from Google predictably. From Hanau for the Rhine-Main region and the German-speaking world.
Media on Earth is an agency for web design, software development and SEO in 63456 Hanau (Karlstr. 63), founded in 2018 by Karim Hadj Yahia. Contact: info@mediaonearth.com, +49 172 6974070.
## Web design
- [Web design](https://mediaonearth.com/leistungen/webdesign/): Websites that load fast, get found on Google and bring in enquiries. Price: from €2,500 (simple site).
- [Website relaunch](https://mediaonearth.com/leistungen/website-relaunch/): New design and faster technology without losing visibility on Google. Also WordPress to Astro. Price: from €2,500.
## Case studies
- [Dogcare24](https://mediaonearth.com/referenzen/dogcare24/): Organic Google search now delivers almost 50 leads per day, more than twice as many as in December 2025.
## Optional
- [About](https://mediaonearth.com/ueber-uns/)
- [Contact](https://mediaonearth.com/kontakt/)You can find the full file, in German, with all services, locations, case studies and guides at mediaonearth.com/llms.txt.
Who actually reads llms.txt?
Hardly any AI service that generates answers for users. On 97% of websites with an llms.txt, nobody requested the file in May 2026. It is mainly read by coding assistants. The figures in detail:
Ahrefs analysed the traffic to 137,210 websites in May 2026 for a study. 28% of them had an llms.txt. On 97% of these websites, the file was not requested a single time during the entire month. The remaining requests came mostly from bots that have nothing to do with AI, such as SEO tools and general crawlers. Among AI bots, GPTBot, the OpenAI crawler, requested the file most often. Bots that fetch pages for live answers in ChatGPT, Perplexity or Claude accounted for only 1.1% of requests. The websites belong to users of the Ahrefs analytics tool and tend to be tech-savvy, which is why the share with an llms.txt is so high. Ahrefs also points out that a request does not mean the content is actually used.
SE Ranking examined around 300,000 domains. Only 10.13% of them had an llms.txt. The study found no link between the file and the number of mentions in AI answers. The authors’ prediction model even became slightly more accurate when they removed llms.txt as a feature.
Google writes in its guide to generative AI in Google Search that Search ignores llms.txt files. Maintaining one is fine, but it neither helps nor harms visibility or rankings. According to Google, no new machine-readable files or special markup are needed to appear in AI Overviews or AI Mode.
The file is mainly read by coding assistants and AI agents that specifically load documentation. In the Ahrefs analysis, Claude Code, a coding assistant from Anthropic, requested llms.txt files more often than any bot for live answers. Accordingly, it is mainly software providers that maintain an llms.txt, for example Anthropic, Stripe and Vercel (as of October 2026).
Is an llms.txt worth it for your website?
For a normal company website, the current data shows no measurable benefit, but it does no harm either. For software providers with documentation, it is worth it.
| Your website | Our recommendation |
|---|---|
| Company website of a service provider or trade business | You can create one, the effort is small. Do not expect more enquiries or mentions in ChatGPT from it. |
| Software, app or API with documentation | Yes. Coding assistants and AI agents use the file to find the right pages. |
| Online shop | Optional. SEO plugins sometimes generate it automatically, in which case maintaining it costs nothing. |
| A provider wants to sell you the file as an “AI visibility package” | Don’t. You can make the file yourself in an hour, and there is no evidence that it affects AI answers. |
We create an llms.txt for every website we build, because it takes hardly any effort and is ready in case more AI services read it in future. We do not sell it as a lever for more visibility.
How to create an llms.txt in five steps
1. Select the important pages
Include only the pages a customer needs to know: services with prices, case studies, locations, contact and your best guides. Legal notice, privacy policy, archive pages and pages with noindex do not belong in it. Our own file comes to around 45 links.
2. Write the file
Create a text file named llms.txt and follow the structure above. Write each description so that it makes sense on its own: what the page offers, for whom and, where applicable, at what price. Name, address and phone number should read exactly as they do on your website, in your Google Business Profile and in directories.
3. Upload and check
Upload the file to the root directory of your domain so that it is available at yourdomain.com/llms.txt. Save it as UTF-8 so that special characters such as German umlauts come through correctly. Then open the address in your browser: you should see plain text, no error page and no download. Also check that all links work.
4. Optional: point to the file
Version 2 of the proposal recommends that pages point to their llms.txt, either with <link rel="describedby" href="/llms.txt"> in the page head or via an HTTP header. Pages can also offer a plain Markdown version and point to it with rel="alternate" type="text/markdown". For company websites this is not a must. We have added the reference to the llms.txt on our website because it is only one line of code.
5. Keep it up to date
An llms.txt with old prices or dead links is worse than none. The simplest option is a file that updates itself:
- WordPress: Yoast SEO has generated the llms.txt for free since version 25.3 (June 2025) and updates it every week. Since version 25.6 you can also choose the pages yourself. You switch the feature on in the Yoast settings.
- Shopify: Since the end of March 2026, Yoast SEO for Shopify also creates the file automatically, with the best-selling products and the largest collections.
- Astro, Next.js and other modern systems: The file is generated from the existing content when the website is built. That is how we do it too: our llms.txt is regenerated with every change from the service, case study and guide pages, including prices. It cannot go out of date because it is not a separate copy of the content.
What an llms.txt cannot do
- It does not block anything. You control which crawlers may read your website via robots.txt. There you can, for example, allow or exclude GPTBot. The llms.txt is only a signpost.
- It does not replace content. What AI systems say about you comes from your website and from what others write about you. A summary in a text file changes little if the pages themselves are thin.
- It does not get you into Google’s AI Overviews. Google ignores the file, see above.
What works for AI visibility instead
According to Google, the same fundamentals apply to its AI features as to normal search: a technically sound website whose content Google can access, and original, expert content that goes beyond general knowledge. From our work on AI SEO, we add four points:
- Content in the HTML. According to an analysis by Vercel from December 2024, the crawlers of OpenAI, Anthropic and Perplexity do not execute JavaScript. Anything that only loads in the browser is invisible to them.
- A first paragraph that answers the question. Every important page should say at the top, in two or three sentences, what you offer, for whom and at what price. Paragraphs like these can be quoted.
- Consistent company details. Name, address, phone number and services should read the same on your website, in your Google Business Profile, on LinkedIn and in industry directories.
- Mentions on other websites. AI systems base their answers on sources. Businesses that appear in specialist articles, directories and comparisons are more likely to be mentioned. According to Google, bought or staged mentions help less than it seems.
The best way to check whether this works is directly: once a month, ask ChatGPT, Perplexity and Google the questions your customers ask, and note who gets mentioned. Google also recommends the performance report for generative AI in Search Console.
Conclusion
Create an llms.txt if you are working on your website anyway. It costs little and does no harm. But according to the current data, it will not bring you new customers: hardly any AI service requests it for its answers, and Google ignores it completely. Your time is better spent on pages that fully answer your customers’ questions. How we do this for businesses is explained under AI SEO. If you want to know where your website stands today: the initial analysis is free.