LLMs.txt generator: build a valid llms.txt in under a minute
An llms.txt file is a Markdown index of your site's most important pages, served at /llms.txt. It is the proposed counterpart to robots.txt and sitemap.xml, designed for AI assistants and inference-time agents that need a curated map of your content rather than a full sitemap dump.
This free generator produces a spec-compliant llms.txt that follows the original llmstxt.org format - H1 site name, blockquote summary, free-form intro, and one or more H2 sections of links with descriptions. Fill in the form, copy the output, and upload it to your domain root.
Honest note: as of April 2026, no major AI provider has publicly confirmed that GPTBot, ClaudeBot, or PerplexityBot reads llms.txt during inference or training. Adoption is still aspirational. We recommend publishing one anyway because it is a low-cost bet on a standard that may stick - and because the structured Markdown is genuinely useful to any agent that does fetch it.
Site basics
Generated llms.txt
# Your Site Name
https://yoursite.com/llms.txt. Serve over HTTPS with status 200. Most publishers serve it as text/markdown or text/plain.One step further
llms.txt only helps if AI bots can read your linked pages
GPTBot and ClaudeBot don't execute JavaScript, so a JS-rendered site shows up empty even with a perfect llms.txt. PageGlass renders your site for bots at the DNS layer - one CNAME, no code changes.
Try free for 7 daysHow to use this tool
- 1
Enter your site name (H1) and a one-sentence summary blockquote
- 2
Add the most important sections of your site - typically Docs, Pricing, Use Cases, and an Optional section (links there are de-prioritised by AI clients)
- 3
For each link, include a full URL and a short description so the AI client knows what each page covers
- 4
Review the live preview on the right - the validator flags any spec issues in real time
- 5
Click 'Copy' or 'Download' to grab the file, then upload it to https://yoursite.com/llms.txt
- 6
Verify it's served correctly by visiting that URL - it should return 200 and the raw text
How it works
The output follows the llmstxt.org specification published by Jeremy Howard in September 2024. The required structure is: a single H1 with the site name, a blockquote summary, optional free-form prose, then one or more H2 sections containing Markdown links with descriptions. The spec also recognises a special H2 section named 'Optional' whose links may be skipped by AI clients with limited context.
The generator validates the file against the spec - H1 present, blockquote present, URL validity per link, duplicate section names, and overall file size. HTTPS URLs are recommended; the validator flags HTTP URLs as warnings rather than errors. The spec doesn't formally limit headings beyond the H1+H2 backbone, so deeper nesting is allowed in practice; the generator focuses on the load-bearing rules.
We do not crawl your site or fetch any external URLs. The tool runs entirely in your browser. The links you enter are formatted into the spec layout and shown in a live preview - nothing is stored or sent anywhere.
Companion convention: many sites publish a separate llms-full.txt that contains the full Markdown content of every linked page concatenated together. This is more useful for ingestion pipelines and IDE assistants than for real-time AI lookups. The generator covers llms.txt only - llms-full.txt is typically built by your docs platform or a documentation generator.
Why this matters for your site
The llms.txt proposal was published by Jeremy Howard at llmstxt.org in September 2024. It addresses a specific problem: an AI assistant trying to answer a question about a website usually cannot fit the whole site into its context window. A curated index of the most important pages, with one-line descriptions, helps the assistant decide which pages to fetch.
Adoption is real but smaller than the hype suggests. Public directories like directory.llmstxt.cloud track tens of thousands of sites publishing llms.txt by late 2025. Major sites with one include Anthropic, Cloudflare, Stripe, Vercel, Mintlify, and Cursor. However, independent audits from third-party SEO firms have found no measurable correlation between publishing llms.txt and improved AI citation rates. Google's John Mueller has stated publicly that Google does not use llms.txt as a ranking signal, comparing its current status to the keywords meta tag, and noted that AI providers have not confirmed they read it.
What reliably matters for AI search visibility is that crawlers can reach and read your content: a clear semantic HTML hierarchy, accessible content (no JavaScript walls in front of the body text), respectful robots.txt directives, and a working sitemap.xml. Structured data (JSON-LD with Schema.org) is good hygiene, but a controlled 2026 study found no citation uplift from it and Google says it is not required for its AI features. AI crawlers vary in JavaScript support: GPTBot and ClaudeBot do not execute JavaScript at all, while PerplexityBot's JS support has been inconsistent in third-party reports. As a result, content rendered client-side by React, Vue, or Angular is at best partially visible and often completely invisible to these crawlers no matter how perfect your llms.txt is. If your linked pages depend on JavaScript to render their content, AI crawlers that do fetch them will often see empty shells.
Frequently asked questions
What is llms.txt and how is it different from robots.txt?▾
llms.txt is a Markdown file at the root of your site that gives AI assistants a curated map of your most important pages. Unlike robots.txt, it does not control crawler access - there are no allow/disallow rules. It is closer in spirit to a sitemap.xml written in Markdown, but with descriptions and curated for inference-time AI clients rather than search engines.
Do AI providers actually read llms.txt?▾
As of April 2026, none of the major AI providers (OpenAI, Anthropic, Google, Perplexity) have publicly confirmed they read llms.txt. Independent audits show AI bots rarely fetch the file. Google's John Mueller compared its current status to the keywords meta tag - aspirational rather than load-bearing, and on 15 June 2026 Google stated that Search ignores llms.txt. We still recommend publishing one because it is low-cost and the standard may yet be adopted.
Where should I host the file?▾
At the root of your domain, exactly like robots.txt. The URL should be https://yoursite.com/llms.txt - not /.well-known/llms.txt. Serve it over HTTPS with a 200 status code. The spec recommends text/markdown as the content type.
What is the difference between llms.txt and llms-full.txt?▾
llms.txt is a small index file with links and descriptions, typically under 10 KB. llms-full.txt is a community convention - not part of the original spec - that concatenates the full Markdown content of every page into a single file, often megabytes in size. llms.txt is for real-time agents with tight context budgets; llms-full.txt is for ingestion pipelines and IDE assistants. This generator builds llms.txt; llms-full.txt is usually produced by your docs platform.
Will publishing llms.txt help my Google rankings?▾
No. Google's John Mueller has stated publicly that llms.txt is not a ranking signal in Google Search and compared its current usefulness to the keywords meta tag. The benefit, if any, is in helping AI assistants find and reference your content more efficiently. For Google rankings, the proven levers are still page experience signals, content quality, structured data, and links.
Can I have headings deeper than H2 in my llms.txt?▾
The spec's load-bearing structure is one H1 (the site name) followed by any number of H2 section headers with Markdown link lists underneath. H3 and deeper aren't part of the core structure, but the spec doesn't strictly forbid them and some publishers (including Mintlify-hosted docs) use H3 for sub-groupings inside an H2. For maximum client compatibility, stick to H1 + H2; if you need finer organisation, prefer splitting content across multiple H2 sections.
Why does my site show up empty when ChatGPT browses to it?▾
Almost certainly because your site is JavaScript-rendered and ChatGPT's crawler (GPTBot) does not execute JavaScript. It fetches raw HTML and stops. An llms.txt full of links to JS-rendered pages just sends the AI crawler to blank shells. Run our Crawler Simulator to confirm what GPTBot actually sees, then prerender or server-render your content so AI crawlers can read it.
Should I include every page in my llms.txt?▾
No. The whole point of llms.txt is curation - it is the index of pages you want AI assistants to use, not a full sitemap. Include the highest-value pages: documentation, pricing, key product pages, primary use cases. Leave out auth-walled routes, marketing campaign landers, and stale content. If you want the full URL list, that is what sitemap.xml is for.
Related free tools
Crawler simulator
Simulate how specific crawlers - Googlebot, GPTBot, Facebook, LinkedIn, and more - see your JavaScript site. Choose a bot and test instantly.
Try it freeWhat bots see
Enter any URL and see the raw HTML that search engine bots, AI crawlers, and social platforms receive - before and after JavaScript rendering.
Try it freeSEO audit
Run a free SEO audit on any website. Check meta tags, headings, structured data, alt text, canonical URLs, and crawler accessibility.
Try it freeMeta tag checker
Test every SEO and social meta tag on any page. Title, description, canonical, robots, viewport, Open Graph, Twitter - with bot-eye-view validation.
Try it freeOG image checker and Open Graph debugger
Test your Open Graph and Twitter Card meta tags. Preview exactly how your links appear when shared on Facebook, LinkedIn, Twitter/X, and WhatsApp.
Try it freeLovable SEO audit
Audit any Lovable site for the JavaScript SEO problem. Lovable apps render in the browser - bots see an empty shell. This tool shows you what's missing.
Try it freeMake sure AI crawlers can actually read the pages you link to
An llms.txt full of links is no use if GPTBot, ClaudeBot, and PerplexityBot can't render your JavaScript. PageGlass renders client-side pages into clean HTML AI bots can read. One CNAME, no code changes.