llms.txt Generator:
Draft It From Your Real Pages.
Enter your domain and get a draft llms.txt built from your actual site - sitemap-discovered pages, real titles, real meta descriptions, grouped into sections by your URL structure - in an editable box with copy and download. Plus an honest read on what your /llms.txt serves today.
How it works
- Discovery - your sitemap first (index sitemaps traversed), homepage links as the fallback. Same-origin pages only, asset URLs skipped.
- Sampling - up to a dozen pages picked round-robin across your URL sections, shallowest first, so every part of the site is represented.
- Robots respected - we fetch as Citevera-Scanner and skip anything your robots.txt disallows for us, reporting the skip count.
- Composition - H1 title from your homepage, blockquote description from your meta description (or a marked placeholder if you have none), then ## sections of Markdown links in the llmstxt.org shape.
- Current-file check - the same homepage-in-disguise detection as the llms.txt reality check, so you know whether publishing the draft is the fix or the serving is.
Example result
A real run of this generator against our own site on 2026-08-04, trimmed to the opening sections:
# Citevera > Be the site ChatGPT, Claude, Perplexity, and Google AI Overviews recommend. Citevera scans your site, fixes what AI engines cannot read or trust, and gets you in the answer. Free scan, no signup. ## Main pages - [Blog](https://citevera.com/blog): Articles on AEO, GEO, llms.txt, schema.org for AI, and making content citable by AI answer engines. - [Free AI Search Tools](https://citevera.com/tools): Free checks that fetch your site the way AI crawlers do: llms.txt serving reality, AI crawler access by user-agent, and more. - [Pricing](https://citevera.com/pricing): Free audit forever. One plan from $49/mo with weekly AI citation monitoring included - no API keys to bring. ## Blog - [Answer engine fanout: one query, ten sources](https://citevera.com/blog/answer-engine-fanout): When a user asks an AI answer engine a single question, the engine decomposes it into several underlying retrievals. - [The 35-point AEO checklist, explained](https://citevera.com/blog/35-point-aeo-checklist): Citevera audits score every site on 35 specific AEO signals organized into six clusters. [... 2 more sections from the 11 sampled pages ...]
Common llms.txt drafting problems
- Dumping the whole sitemap - an llms.txt is a curated index; engines want your best entry points, not everything.
- Descriptions written for humans mid-journey ("Learn more here") instead of saying what the page answers.
- A missing one-paragraph summary - the blockquote under the H1 is the highest-leverage text in the file.
- Links to pages that later move - regenerate after restructures, or automate regeneration.
- Publishing the file but never fixing serving - on WordPress plain permalinks, /llms.txt can return your homepage no matter what you upload.
- Writing it once and forgetting it - the file drifts from the site with every post you publish.
Frequently asked questions
How is this different from llms.txt templates and paste-in builders?
Templates give you placeholder structure to fill by hand. This generator crawls a sample of your actual site - sitemap first, homepage links as fallback - and drafts the file from your real page titles and meta descriptions, grouped into sections by your URL structure. What you edit is already true; nothing is invented.
How many pages does it use?
Up to a dozen, chosen for section variety (round-robin across your URL sections, shallowest paths first) so a large blog cannot crowd out your pricing or product pages. An llms.txt is a curated index, not a sitemap dump - a dozen well-chosen links beats five hundred.
Does the crawl respect robots.txt?
Yes. We fetch as Citevera-Scanner and honor your robots.txt for our own token: if you disallow us on a path, that page is skipped and the result says how many were skipped. If you disallow us site-wide, we stop and tell you - we do not crawl around your rules.
What does the current-file comparison show?
The same detection our llms.txt reality check uses, run on what your /llms.txt serves today: no file, a real file (with word and link counts), or the WordPress plain-permalinks case where the URL answers 200 with your homepage in disguise. Publishing a draft does not help until the serving problem is fixed, so we surface it here.
Should I publish the draft as-is?
No - edit it first, and the draft says so. Page titles are written for browser tabs, not for AI navigation; the highest-value edit is rewriting the one-line descriptions to say what each page answers. The structure, URLs, and section grouping are the part worth automating.
Format background: how to generate llms.txt, writing llms-full.txt, and what llms.txt is. Already have a file? Run the reality check on it.
Built by Paul, founder of Citevera · Published 2026-08-04 · Updated 2026-08-04 · Pairs with the Citevera WordPress plugin, which regenerates llms.txt automatically on every save