Free llms.txt Generator
Enter a domain. We read the site’s sitemap and the titles and descriptions it already publishes, and turn them into an llms.txt in the llmstxt.org format. Nothing is rewritten by a model. If a file exists, it is checked against the format first.
Check whether AI crawlers can read your site
A file is only useful if the crawlers can fetch it. The crawler checker reads your robots.txt and firewall the way they do.
What llms.txt is
llms.txt is a plain Markdown file at the root of a site, next to robots.txt, that gives a language model a short map of what the site is and where its important pages are. The format, published at llmstxt.org, is deliberately small: an H1 with the site name, an optional one-line summary in a blockquote, optional paragraphs of detail, and then H2 sections that each hold a list of links written as "- [title](url): note". A final section named Optional holds links a reader can skip when it needs a shorter context.
That is the whole specification. There is no schema, no required sections, and no rule about what to link. The file exists because a model with a limited context window does better with a curated list of 40 pages and one-line notes than with a 5,000-URL sitemap and no notes at all.
What it does and does not do
Be clear-eyed about this. As of 28 September 2026, none of the AI search vendors documents reading llms.txt. OpenAI’s crawler documentation, Anthropic’s crawler support article and Perplexity’s bots page describe robots.txt and say nothing about llms.txt. Google has said publicly, through Gary Illyes and John Mueller, that Google does not use it and has no plans to. Each of those companies does publish an llms.txt for its own developer documentation, which tells you they expect coding tools and agents to read such files, not that their answer engines fetch yours.
So the file is not a ranking signal and not a substitute for being crawlable. What it is: a cheap, standard, human-readable index that any agent, tool or model given a URL can use, and a discipline that forces you to write one clear line about each important page. If a vendor starts reading it, you are ready. If one never does, you have lost fifteen minutes.
Installing it by platform
The result includes step-by-step instructions for the platform it detects, checked against each platform’s own documentation. In short: Shopify has served /llms.txt on every store since May 2026 and you replace it with a templates/llms.txt.liquid file in the theme code editor. WordPress sites either turn on the LLMS.txt feature in Yoast SEO, which generates its own version, or upload this file to the web root. Webflow takes an upload under Site settings, SEO. Squarespace 7.1 has an LLMS.txt tab under SEO Settings. Wix generates one automatically on upgraded sites with a custom domain and lets you edit it under SEO & GEO. Anything else: put the file at the root, next to robots.txt.
Two platforms cannot serve a root text file by themselves: Squarespace 7.0 and Wix sites that do not meet the requirements above. For 7.0 the workaround is a file upload plus a URL mapping; for Wix there is no workaround short of upgrading.
llms.txt versus robots.txt versus sitemap.xml
The three files answer different questions. robots.txt says who may crawl what, and every major crawler reads it; it is the only one of the three that can keep a bot out. sitemap.xml says which URLs exist and when they changed, for crawlers that index everything; it is exhaustive and has no descriptions. llms.txt says which pages matter and what each one is about, for a reader that will not crawl everything; it is selective and made of prose.
They do not replace each other. A site that blocks OAI-SearchBot in robots.txt is invisible to ChatGPT search no matter how good its llms.txt is, which is why the generator links to the crawler checker below the result. And a site with no sitemap gives this generator much less to work with: it falls back to the links on the homepage.
Why we use your own descriptions instead of AI rewrites
Every note in the generated file is the site’s own meta description, product description or collection description, stripped of HTML and cut at the first sentence or 160 characters. No model rewrote anything. That is a deliberate choice. A rewrite can drift from what the page says, and a file that misdescribes a page is worse than no file, because a model that trusts it will misquote you. Your descriptions are also what your pages already show search engines, so the file and the site agree.
The side effect is useful: pages with no meta description show up with no note, and a homepage with no meta description produces a file with no summary line. Both are worth fixing on the page, where every crawler benefits, rather than in the file alone.
Shopify stores
On a Shopify store the generator switches to a store-shaped file: collections with their descriptions, the first fifty products in sitemap order with one-line notes, the policy pages that exist, the about and contact pages, and the blogs under Optional. It also recognises the default file Shopify serves and shows what the generated version adds. The Shopify page explains how the pieces map and how to install through the theme template.
Questions
Will an llms.txt get my site into ChatGPT or Perplexity answers?
Not by itself. None of the AI search vendors documents reading the file, and Google has said it does not use it. What gets a page cited is that crawlers can reach it and that it says something quotable. The file is a low-cost index for tools and agents that do read it.
How does the generator choose which pages to include?
It reads the sitemap, groups URLs by their first path segment, and fetches up to 30 pages, starting with the homepage sections and pages like about, pricing, contact and docs. Each fetched page becomes a link with its title and meta description. Sections with more pages than were read get a single link under Optional with the count.
What does the linter check on my existing file?
That the H1 comes first, that there is a blockquote summary, that every list item is a "[title](url)" link, that no H3 or second H1 appears, that sections are not empty, that the file is under 100 KB, and that a sample of its links still return a page.
Why did it say the run stopped early?
Each run may make at most 45 requests to your site and has fifteen seconds. Large sites hit one of those limits. The file you get still covers every section the generator saw; it lists fewer individual pages and summarises the rest.
Do you store the file or my domain?
Results are cached for six hours per site so a shared link loads instantly, and each connection gets ten fresh runs an hour across the free tools. There is no account and no email step.
Can I edit the file before installing it?
Yes, and you should. Reorder sections, remove pages you would rather not highlight, and tighten notes. Keep the structure: one H1, one blockquote, H2 sections of link lines.