# llms.txt Generator > Create llms.txt from a sitemap: fetch each page's title and description, group by path, exclude URLs, cap the count, and serve it with the right Content-Type. - URL: https://bytescope.dev/en/llms-txt-generator/ - Locale: en · Also: [ja](https://bytescope.dev/ja/llms-txt-generator.md) · [zh-Hant](https://bytescope.dev/zh-Hant/llms-txt-generator.md) - Runs: Sitemap URLs and page heads are fetched by our Worker and not stored; an uploaded XML file stays in your browser. - Requires: Any modern browser (Chrome, Edge, Firefox, Safari), desktop and mobile. - Updated: 2026-09-08 ## What it does `llms.txt` is a Markdown file at your site root that gives AI tools a short, curated map of your pages: an H1 with the site name, a one-line summary in a blockquote, then H2 sections of `- [title](url): description` links (the llmstxt.org convention). This tool builds one from your sitemap, so you do not have to type hundreds of lines by hand. ### How it works Paste a sitemap URL or upload the XML. A sitemap index is followed automatically, up to 20 child sitemaps. Pages are grouped into sections by the first path segment, a group with only one page folds into the top section, and every section can be renamed. If your URLs start with a language prefix such as `/en/` or `/ja/`, pick one language so the file is not three copies of the same site. Exclusion rules use robots.txt pattern syntax: `*` matches anything and a trailing `$` anchors the end. The page cap defaults to 500 and goes up to 2,000. ### Titles and descriptions Fetching titles reads each page's `` and `<meta name="description">` through our Worker, 10 pages per request with at most 2 requests in flight. The Worker allows 60 requests a minute per IP address, so a 500-page site takes about a minute, and a rate-limit pause is waited out automatically. You can stop at any point and keep what has arrived, or skip the step and use each URL's path as its title. Nothing is written by an AI: the descriptions are the ones already on your pages. ### Where to put it Serve the file at `/llms.txt` on your site root with `Content-Type: text/plain; charset=utf-8`. Most hosts already send `text/plain` for `.txt`, but some leave out the charset, which garbles non-English titles, and a single-page app often answers unknown paths with its HTML shell, which an AI tool then reads as your llms.txt. The snippets under the preview set the header on Cloudflare, Netlify, nginx and Vercel. ## Limits Does not write page summaries with an LLM: it uses each page's own title and meta description. Sitemaps and page heads are fetched through our Worker, 2,000 pages at most, and nothing is stored. ## FAQ ### What is the llms.txt format? A Markdown file proposed at llmstxt.org: one `#` heading with the site or project name, an optional `>` summary, then `##` sections listing links as `- [title](url): notes`. A section named `Optional` marks links an AI tool can skip when its context is short. ### How do I create llms.txt from a sitemap? Paste the sitemap URL (usually `/sitemap.xml`, or the `Sitemap:` line in robots.txt), pick a language if the site has several, add exclusions, fetch titles, rename sections if you like, then download the file and upload it to your site root. ### Which Content-Type should llms.txt use? `text/plain; charset=utf-8`. The content is Markdown, but plain text is what every client accepts, and the charset keeps non-English titles readable. After deploying, run the AI search check on your homepage: its llms.txt card flags a file that comes back as HTML. ### Does it use AI to write the descriptions? No. Each link uses the page's own `<title>` and meta description, cleaned of line breaks and extra spaces. If a page has no description the link has no notes. Edit the file afterwards if you want a better summary. ### Why are some titles just URL paths? That page's head could not be read: it returned an error, redirected somewhere that failed, was not HTML, or has no `<title>`. It also happens for every page when you skip the fetch or stop it early. The path is a readable fallback, not an error in the file. ### Is my sitemap or site stored? No. An uploaded XML file is parsed in your browser. A sitemap URL and the page heads are fetched by our Worker, which returns them to your browser and keeps nothing, with 60 requests a minute per IP address. ## Related tools - [AI Search Check](https://bytescope.dev/en/ai-search-check.md) — Check whether AI search can fetch, read and cite a page, llms.txt included. - [AI Crawler Log](https://bytescope.dev/en/ai-crawler-log.md) — See which AI crawlers visit, and which sitemap pages they never fetched.