Create your llms.txt file with AI

What is llms.txt?

llms.txt is a plain-text summary that introduces a website's key content to large language models (LLMs) in a single file. The proposal was published in September 2024 by Jeremy Howard, co-founder of Answer.AI. It started from a practical problem: HTML pages are cluttered with menus, scripts and ad code, while the amount of text a model can process in one go, its context window, is limited. To get around this bottleneck, the file presents the essence of the site at a single address in Markdown, a format built on headings and list markers.

The tradition of bot files in the root directory, the site's main folder, began with robots.txt in 1994; llms.txt brings the same approach to AI bots. The full definition of the standard is published at llmstxt.org, and its scope is deliberately narrow: the site name, a short summary describing the brand and an annotated list of key pages. The file is reached the same way on every site: at /llms.txt in the root of the domain.

Which AI engines actually read the summary file?

Official support lags behind the proposal's popularity: as of our review date (July 22, 2026), no major AI engine had officially confirmed using llms.txt when generating answers. Google takes the clearest position: John Mueller of the search team wrote in April 2025 that no major AI service had said it uses the file, and compared it to the keywords meta tag that search engines have long ignored. Mueller also said that a site owner looking at server logs could see for themselves that bots do not fetch the file. Claims of support for OpenAI and Perplexity circulating on third-party blogs could not be verified against official documentation in the same review.

Model makers' own behavior paints a mixed picture: Anthropic and Mistral each publish an llms.txt for their developer documentation, yet neither has announced that its bots consume the files on other sites when generating answers. Publishing and consuming are different things, and support lists that confuse the two overstate adoption. Today the verifiable use of the format is mostly on the developer side: code tools such as Cursor use these files when feeding documentation content to a model.

Which sections does the file contain?

The standard structure has only one required element: an H1 heading with the site's name. The remaining sections are recommendations and appear in this order:

• Blockquote summary: one or two sentences explaining what the brand does; this is the main sentence the model relies on when describing the site.
• Free-form notes: short context paragraphs such as target audience and service scope.
• Link lists grouped under H2 headings: each line has a page title, a URL and a one-line description (services, guides, pricing pages).
• The Optional section: secondary pages the model can skip when context is tight.

Two mistakes are common in practice. If an advertising slogan goes into the summary line, the model will describe the brand with that slogan; a plain, factual definition always produces a more accurate result. The second mistake is filling the file like a sitemap: a list with hundreds of URLs stops being a summary, because the point of the format is to highlight a small number of selected pages.

How does it differ from robots.txt and sitemap.xml?

The three files sit in the same root but answer three different questions. robots.txt uses rule lines (User-agent, Disallow) to tell bots where they may not go; sitemap.xml lists the URL inventory in XML format; llms.txt summarizes what the content is about. The age gap is large too: Google announced the sitemap protocol in 2005, while the summary proposal is less than two years old.

In practice, access comes first: if GPTBot (OpenAI) or ClaudeBot (Anthropic) is blocked in robots.txt, the summary file in the root is never read. You can test which AI bots can reach your site in a few seconds with the AI crawler audit tool; the summary file only makes sense after this check. The three are complementary, not competing; in a solid setup, robots.txt, sitemap.xml and llms.txt sit side by side in the root.

How do you create llms.txt and where do you upload it?

You can prepare the file in two ways: write it by hand in a text editor following the llmstxt.org template, or leave it to the tool on this page. The tool works in three steps:

1. Enter your website URL; the tool scans your homepage and sitemap and collects page titles.
2. From this scan, AI compiles the site description, a service summary and a list of key pages in Markdown.
3. Copy the output in one click or download it as llms.txt.

Upload the generated file to your site's root directory; the right location is wherever yourdomain.com/llms.txt opens from. On sites using cPanel this is usually the public_html folder; on WordPress, the file goes into the same folder via FTP or the file manager in your hosting panel. To check, open the URL in a browser: if you see plain text, the setup is complete. One exception is often missed: each subdomain uses its own root, so blog.yourdomain.com needs a separate file.

A site owner selling services keeps the list short: service pages, pricing, contact and about go into the main list, and the blog archive moves to the Optional section. The file is not uploaded once and forgotten; when a new service page launches or prices change, the list is updated by hand.

What is llms-full.txt for?

llms-full.txt is the extended version that combines the full text of selected pages into a single Markdown file. The division of labor is clear: llms.txt provides the map, llms-full.txt carries the content itself. Its most common use is on documentation sites; Anthropic publishes its developer documentation in both formats, and code tools feed this single file to the model as context.

A full-text file grows quickly; as content increases, it exceeds the model's context window and the end of the file gets cut off. For a small service site, llms.txt on its own is enough; a SaaS product with hundreds of documentation pages, on the other hand, gets value from llms-full.txt to hand its technical content to tools in a single file.

What does llms.txt change in practice?

Four findings stand out. Setup takes minutes and carries no known risk, which makes the file a low-cost preparation step. Official support is weak; expecting guaranteed citations from the file before major engines confirm they consume it is a miscalculation. What really determines visibility in AI answers is content quality and brand signals; the summary file only lays the groundwork for the brand to be described correctly. For a site owner budgeting for AI visibility, the order is clear: bot access and content first, then this file. Regularly monitoring how your brand is mentioned in AI engines with our AI visibility service is the natural next step after this setup.