Free llms txt generator: draft an llms.txt file from your own sitemap

How do you use the llms txt generator?

Type your domain into the box above, press Generate and wait while the tool reads your site. When it finishes, the draft appears in the text area, where you can edit it, copy it or download it as a file named llms.txt.

Behind the button, the process follows a fixed order so the output is predictable. Knowing that order helps you understand why some pages appear and others do not.

• It requests your robots.txt and looks for Sitemap lines, using up to two of them; if none are listed, it tries /sitemap.xml.
• If the sitemap is an index file, it opens the first two child sitemaps and skips image, video and news sitemaps.
• If no sitemap can be found, it falls back to the links on your home page.
• It sorts the URLs it found by depth, shallowest first, and keeps the first 25, so top-level pages win over deep archive pages.
• For each page it reads the title tag, the meta description (or the Open Graph description if the meta is empty) and the first H1.
• It groups pages by the first folder in their URL, so /blog/ pages land under a Blog heading and single-level pages under Main pages.

What does a generated llms.txt file contain?

The draft follows the structure set out in the proposal at llmstxt.org: one H1 with the site name, a blockquote summary, then H2 sections that each hold a Markdown list of links. Every link line carries a page name, the URL and a short note on what the page covers.

The generator fills in most of that for you, using the data your pages already publish. The table shows where each part comes from and what you should check.

The proposal also allows an H2 called Optional for secondary links that a reader with little context space can skip. The tool does not create that section, because only you know which pages are secondary; add it by hand if you want one.

How should you edit the draft before you publish it?

Cut it down, then sharpen what is left. A draft built from a sitemap reflects how your site is stored, not how a reader should meet it, so expect to delete lines, rename sections and rewrite notes.

Keep the pages that answer real questions about your business: what you sell, who it is for, pricing or how pricing works, documentation, policies and contact details. Remove tag archives, author pages, thin category listings, thank-you pages, login screens and anything you would not want quoted. If an important page is missing because it sits deep in the sitemap, add its line by hand.

Specific notes are the main thing worth your time. A note such as "Learn more about our services" tells a model nothing, while "Fixed-scope onboarding for teams of 10 to 200, with a sample timeline" says what the page holds. Many meta descriptions are written to win clicks rather than to describe, so rewrite the ones that read like adverts.

• Rewrite the summary so it states what the business does, for whom and where, in plain words.
• Rename sections to match how a customer thinks: Products, Pricing, Docs, Policies, Company.
• Check that every URL returns a normal page, not a redirect chain or an error.
• Make sure each linked page shows the same facts as the file, including names, prices and locations.

Where do you upload llms.txt on WordPress, Shopify, Webflow or a static host?

Upload it to the root of your domain, so the file opens at yoursite.com/llms.txt as plain text. The exact route depends on how your site is hosted, and platforms change their options, so check your platform's current help pages before you start.

On a self-hosted WordPress site, the usual route is the same one used for other root files: place it in the site's top-level folder through your host's file manager or SFTP. On hosted platforms such as Shopify and Webflow, file access works differently, so look for the platform's own instructions for serving a file at the root, or ask your developer what is possible on your plan.

On Cloudflare Pages and other static hosts, put llms.txt in the folder that your build publishes as the site root, often called public or dist, and deploy. Afterwards, open the URL in a private browser window to confirm it loads, and check that your CDN or firewall rules do not block or challenge requests to it.

What can an llms.txt file not do?

It cannot make an answer engine cite you, because none of the major engines has said it reads the file. Search Engine Land reported that Google's Gary Illyes said Google does not support it and that John Mueller likened it to the meta keywords tag. Google's own AI optimization guide, updated in July 2026, says no special files, markup or Markdown are needed for its AI search features.

It also replaces nothing you already rely on. Crawlers still need pages they can fetch as server-rendered HTML, robots.txt still decides which bots may visit, and XML sitemaps still help search engines discover URLs. If any of those are broken, a tidy Markdown index will not compensate.

That is why we describe the file to clients as optional hygiene. It takes an hour, it does no harm when kept accurate, and it should never be the reason anyone pays for ongoing work.

Who does an llms.txt file actually help?

Mostly developers and the coding agents they use. When documentation is laid out as a clean index of reference pages, an assistant inside an editor, or a developer pasting docs into a chat, can find the right page with less noise.

The companies building these tools publish files like this for their own documentation. OpenAI, for instance, publishes its commerce documentation for merchants as an llms-full.txt file, which places the full text in one document rather than a list of links.

If you run an API, an SDK, a developer platform or a large help center, publishing a well-edited file is a reasonable courtesy to those users. For a brochure site, a local service firm or a small store, the honest expectation is a small, unmeasured benefit.

What actually moves AI visibility if llms.txt does not?

The work that engines document and that you can measure. Start with AI crawler access: allow OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot and Bingbot in robots.txt, check Cloudflare's AI crawler settings and confirm in server logs that those bots get HTTP 200 responses. Our technical AEO service covers that layer, alongside rendering, sitemaps and schema markup.

Next comes the content itself. Pages that open each section with a direct answer, use question-style headings and show a last-updated date are easier to quote, which is the focus of our answer-first content work. Consistent company facts across Organization schema, Wikidata, review sites and LinkedIn matter too, and so do genuine mentions on Reddit and YouTube.

Then measure. We run a fixed prompt set in ChatGPT, Perplexity, Google AI Overviews and Copilot before and after changes and report citation share and AI referrals in GA4 each month. If you want to know where you stand first, an AEO audit records which prompts cite you today and ranks the fixes; if you prefer to work through it yourself, our AEO checklist sets out the same steps in order.