What Is llms.txt and Does Your Website Need It?

What is llms.txt, in plain terms?

llms.txt is an ordinary Markdown text file that lives at yoursite.com/llms.txt. In it you explain in your own words what the business does and list links to the pages that matter most for understanding the company: services, prices, contacts, answers to common questions.

The idea is simple. A typical web page is cluttered with menus, banners, scripts and pop-ups. A language model has an easier time reading a short summary than parsing the layout, and llms.txt puts that summary in one place.

The format is described at llmstxt.org. The same proposal includes llms-full.txt, a variant that holds the full text of your key pages in a single file instead of just links.

How is llms.txt different from robots.txt and sitemap.xml?

All three files sit in the site root, and people mix them up all the time. Each has its own job.

robots.txt controls access, sitemap.xml gives the full list of URLs, and llms.txt sets the priorities. If AI crawlers are blocked in robots.txt, llms.txt won't help, because nobody will reach it. So check these files together.

Do ChatGPT, Gemini and Perplexity actually read llms.txt?

The honest answer: there's no guarantee. OpenAI, Google and Perplexity haven't publicly committed to using llms.txt when generating answers. ChatGPT with search relies in part on the Bing index, Gemini and AI Overviews rely largely on Google, and regular site pages remain their main source.

That said, AI agents and tools that are given a link to your site can read the file just like any other page. The format is young: support may grow, or it may stay where it is today.

We treat llms.txt as cheap insurance. It doesn't replace content, markup or external mentions, but it costs next to nothing.

How do you write llms.txt? File structure

The format is flexible, but there's an established order of blocks.

Stick to facts. “The best service in town” tells a model nothing. “A dental clinic in Kyiv offering implants and orthodontics, with appointments in Ukrainian and English” gives it several verifiable facts at once, and that's the kind of sentence AI assistants tend to repeat.

Which pages should go into llms.txt?

Include the pages an AI assistant should draw on when it answers questions about you. For most businesses that means:

• the home page and an About page with your story and team;
• service or product category pages;
• pricing, or at least starting prices;
• answers to frequently asked questions;
• contact details: address, phone, opening hours, cities and countries you serve;
• case studies, guides and articles that answer customer questions.

Leave out the cart, account area, filter pages and technical sections. Links should point to live pages that return a 200 status code, with no redirects. Keep each link description to one sentence: it helps a model understand what's on the page.

How do you publish the file and check that it works?

1. Save the file as llms.txt with UTF-8 encoding.
2. Upload it to the site root so it opens at yoursite.com/llms.txt.
3. Open that URL in a browser: you should see plain text, not a 404 page or a redirect to the home page.
4. Check robots.txt: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Google-Extended must not be blocked.
5. Update the file whenever your services, prices or contact details change.

On WordPress and custom-coded sites, the file goes up easily through hosting or a plugin. On Wix, Tilda and Squarespace, access to the site root is limited, and uploading the file isn't always possible. That's one reason site builders are only a partial fit for GEO.

Does your site specifically need llms.txt?

Almost every site should have one: the effort is minimal and there's no risk. It's especially useful if your site is heavy, script-driven, has many pages or runs in several languages. A short summary helps any reader get the point fast.

Don't expect miracles, though. On our own site, insync.com.ua, llms.txt is part of the GEO setup, and after that setup went live, citations of the site in ChatGPT went from 0 to 24 in about a week and a half. There's no way to isolate the file's share of that result: content, markup, open crawler access and llms.txt all worked together. For category queries like “best agency…”, we still aren't named, because external mentions decide those.

To see what else drives visibility, read our guides what GEO is and how it differs from SEO and why ChatGPT recommends your competitors instead of you.

How do you keep llms.txt up to date without manual work?

Writing the file once is easy. Keeping it current is harder: adding new pages, removing deleted ones and making sure robots.txt, sitemap.xml and Schema.org markup don't contradict each other.

GEO Autopilot automates this work: in the first month we set up the technical foundation, llms.txt included, and then maintain it as new pages go live, 4 of them each month. You can start with the free audit in our Telegram bot: in 60 seconds it checks 8 factors, llms.txt among them, and shows what's missing. If you'd rather look at everything yourself first, follow our step-by-step guide how to check whether ChatGPT sees your business.