LLMs.txt Generator

An llms.txt generator builds the llms.txt file itself, so you don't have to write the Markdown by hand or memorize the format's rules. You provide a site name and a short summary, then add the pages you want an AI assistant to see first - either by importing them straight from your XML sitemap or typing them in manually - and this tool assembles a properly structured, validated llms.txt file you can copy or download.

The file itself, llms.txt, is a proposed convention: plain Markdown placed at a site's root (yourdomain.com/llms.txt) that gives AI assistants and answer engines a short, curated map of a site's most important pages - a heading, a one-line summary, and links grouped into sections such as Docs, API Reference, or Blog. It exists because most sites are too large, too script-heavy, or too cluttered with navigation for a language model to usefully read in full within one request.

Use this generator if you run a documentation site, a SaaS product, an open-source project, or a content site and want an AI-readable index of your best pages without tracking the syntax rules yourself. It imports your existing pages, helps you organize them into sections, and checks the result against what the format expects before you publish it.

The tool is split into two parts - your steps on the left, the live result on the right. Nothing is sent anywhere except the one-time sitemap lookup in Step 2, and even that step is optional.

1. 1
   Add your site details

   Enter your site name and a one-line summary - these become the required H1 heading and blockquote at the top of the file. Add optional extra details if there's useful context worth including.
2. 2
   Auto-import pages, or add them manually

   Enter your website URL and click Import Pages to pull in your XML sitemap's URLs, each with a suggested title and section - review the list, pick the ones worth including, and add them to the builder. Or skip straight to Step 3 and add links by hand.
3. 3
   Organize sections and links

   Rename sections, edit link titles and descriptions, and remove anything that doesn't belong. Empty sections are simply left out of the generated file, and the section list scrolls on its own once it gets long, so it never crowds out the rest of the page.
4. 4
   Watch validation, then copy or download

   The right side updates instantly as you edit. A Valid banner means no issues; otherwise it lists exactly what to fix. Copy or download the finished llms.txt and upload it to your site's root directory.

Each row in the Validation Results panel is color-coded by severity - these are checks against the llms.txt format itself, not a judgment on your content:

• Green (Passed) - a check found nothing wrong, such as "site name set as the H1 heading." No action needed.
• Amber (Warning) - a best-practice recommendation, like a missing summary or a link with no title. Not a broken file, but usually worth fixing before you publish.
• Red (Error) - a genuine problem: no site name, no links at all, or a link missing a valid http(s):// URL. Fix these before uploading the file.

Say a small SaaS company wants an AI coding assistant to find its setup guide and API reference first, instead of wading through marketing pages to get there.

It's worth being precise about what llms.txt can and can't influence:

• Not a direct ranking factor. No search engine has described llms.txt as influencing search ranking position - it has nothing to do with Google, Bing, or any traditional crawler-based index.
• A best practice, not a guarantee. Unlike robots.txt or sitemap.xml, no major AI lab has formally committed to reading llms.txt. Some AI browsing agents and documentation tools already check for it; others don't yet. Publishing one is a low-cost bet on where AI-assisted browsing is heading, not a guaranteed outcome.
• An indirect signal for AEO/GEO, not classic SEO. For a crawler that does read it, a well-curated llms.txt reduces the guesswork of figuring out a site's most important pages from a partial crawl, which can make citations and summaries more accurate. See a full AudEsto audit for how answer-engine optimization is scored more broadly.

• SaaS and developer-tool docs - a Docs and API Reference section pointing an AI coding assistant straight to the pages that actually explain your product, instead of it guessing from scattered marketing pages.
• Open-source projects - linking the README, install guide, and API docs so a model explaining "how do I use this library" cites the current, correct pages.
• Content publishers and blogs - a Blog section highlighting cornerstone articles, so an AI summarizing a topic your site covers well is more likely to find and cite them.
• E-commerce and SaaS pricing pages - a Pricing section pointing directly at your current plans, so an AI answering "how much does X cost" isn't working from a stale cached page.

• Dumping the entire sitemap in unedited. llms.txt is meant to be a curated, high-signal selection, not a second sitemap.xml - importing 500 pages verbatim defeats the purpose.
• Skipping the summary. The blockquote right after the H1 is often the only context a model reads before deciding whether to follow any links at all - leaving it blank wastes the file's most valuable line.
• Leaving every link description blank. Titles alone (especially auto-imported ones from URL slugs) are often ambiguous - a short description disambiguates what a link actually leads to.
• Forgetting to upload it to the root. Same rule as robots.txt - it must live at yourdomain.com/llms.txt, not in a subfolder, or nothing will find it.
• Letting it go stale. A file pointing at a removed product page or an old pricing URL is worse than no file - revisit it after major site changes.
• Confusing it with llms-full.txt. That's a separate, much larger companion file containing full page content, not a link index - see the FAQ below.

• Curate hard. A dozen genuinely important pages beat a hundred mediocre ones - treat this like a README's table of contents, not an index.
• Write the summary like you're briefing a new hire. One clear sentence about what the site is and does, in plain language.
• Put your most important section first. Nothing in the format enforces an order, but a model working with a limited context window is more likely to fully read the first section and only skim or truncate what comes after - don't bury Docs or API Reference under three sections of legal pages.
• Use the "Optional" section name for anything skippable. The convention treats a section literally named "Optional" as safe to drop when a model has limited context to spend - a good place for legal, careers, or other secondary pages.
• Re-run the import after big site changes. A redesign, a new docs section, or a URL migration is worth reflecting here, the same way you'd refresh a sitemap.

• robots.txt - controls which crawlers may request which paths, including AI training bots like GPTBot and Google-Extended. Build one with the Robots.txt Generator.
• Meta tags & Open Graph - per-page signals that shape how a link is titled, described, and previewed. Build them with the Meta Tags Generator and Open Graph Tester.
• AEO / GEO (Answer & Generative Engine Optimization) - the broader discipline of making content easier for AI systems like ChatGPT, Claude, Gemini, and Perplexity to find, understand, and cite. Check your site's AEO/GEO score with a full AudEsto audit.
• Schema.org structured data - machine-readable markup embedded in your pages themselves, complementing an off-page file like llms.txt. Build it with the Schema Markup Generator.

• llms.txt is a voluntary convention - no crawler or AI system is required to read it, and this tool cannot guarantee any specific AI product will fetch or use the file you publish.
• Like robots.txt, the file must live at the domain root - /llms.txt, not /folder/llms.txt - to be found by anything that does check for it.
• Auto-import relies on your XML sitemap being present and correctly declared - a site with no sitemap will need pages added manually in Step 3.
• This tool checks the file's own format and completeness, not whether it will actually change how any specific AI product treats your site - that depends entirely on adoption decisions outside AudEsto's control.

Only the sitemap auto-import step in Step 2 sends anything to AudEsto's servers - the website URL you enter, looked up through the same SSRF-hardened fetcher used by every other AudEsto utility, purely to read your public sitemap. Every other part of this tool - your site name, summary, sections, links, the live preview, validation, copy, and download - runs entirely in your browser and is never transmitted anywhere. Refreshing or closing the tab clears everything, since nothing is stored remotely - copy or download the file before you navigate away if you want to keep it.

Ready to build yours? Scroll up, add your site details, and let AudEsto's llms.txt generator do the formatting.