Free LLMs.txt Generator
Generate an AI-readable site guide that follows the llmstxt.org spec: enter your site name, a one-line summary, and your key pages, and get the file in your browser in about 3 minutes. Built for site owners, content teams, and indie developers.
By the WalkGEO content team Last updated: Content standards
Save the generated content as llms.txt in your site root.
llms.txt is a Markdown file at a site's root that tells AI agents, in a few hundred words, which pages are worth reading. Jeremy Howard of Answer.AI proposed it in September 2024; it reached v2 in August 2026 (source: llmstxt.org). The problem it solves is concrete: a whole site's HTML does not fit in any model's context window, and stripping navigation, ads, and scripts is hard — a hand-written guide fits in one go. According to the v2 proposal, thousands of sites now publish one, and OpenAI, Anthropic, and Gemini each ship one for their developer docs.
The spec is short. In order: an H1 with the site or project name — the only required part; a blockquote summary; any number of plain paragraphs; and H2 sections holding file lists. Each list item is "- [Title](URL): notes", with the notes optional. By convention there is also an ## Optional section for secondary links an agent can skip when context is tight. The file can live at the root or at any subpath: /docs/llms.txt covers only /docs/, and when several exist the agent should use the most specific one. That is more flexible than robots.txt, and it works for GitHub Pages setups where you only control a subdirectory.
How we tested the output format: we put the generator's result next to the mock example on llmstxt.org and the real llms.txt that the FastHTML project publishes, and checked that the three parts — H1, summary, H2 file lists — line up. It is written down because spec compliance is verifiable; you do not have to take our word for it.
What the output looks like
Here is the full file this generator produces for walkgeo.com — real output, not a mock-up. All four structural parts the spec asks for are present: H1, summary, H2 sections, and links with notes.
# WalkGEO
> A GEO agent for AI search: tracks brand mentions and citations across Doubao, DeepSeek, Qwen, and Yuanbao.
## Pages
- [Home](https://walkgeo.com/en): What the product does and which engines it monitors
- [Pricing](https://walkgeo.com/en/pricing): Three plans compared by quota
- [Free tools](https://walkgeo.com/en/tools): Single-page GEO check and llms.txt generator
- [What is GEO](https://walkgeo.com/en/blog/what-is-geo): Definition, how it differs from SEO, how citations work
- [AI citation tracking](https://walkgeo.com/en/blog/ai-citation-tracking): How to check citations and which tools to use
Each H2 section is a file list. The note after the colon says what a reader will find — that is what lets an agent decide whether to follow the link at all.
Five checks for an llms.txt that actually helps
Only the H1 is mandatory, so passing the spec is easy and being useful is not. These five follow llmstxt.org's own writing guidance, and they are the order the generator uses.
-
1
Make the H1 the site name, not Home or Welcome
The H1 is the file's identity anchor — it is how an agent ties the file to a site. Writing Home throws that anchor away.
-
2
Say who you are, what you do, and for whom, in one line
The blockquote is the first thing an agent reads. Copying a slogan is useless; write one fact that stands on its own.
-
3
Keep the list to 10–30 links, ordered by importance
Dumping every URL turns llms.txt into a renamed sitemap. Curation is the value: product, pricing, core docs, your best articles.
-
4
Add a short note to every link
A link followed by ": three plans compared by quota" beats a bare link — the note lets an agent judge relevance without fetching.
-
5
Move secondary links into the Optional section
By convention an agent may skip the whole ## Optional section when context is tight. Archive posts and legacy docs go there so the main list stays short.
How much it actually helps: an honest read
First, what it is not. Google has said plainly that it does not use llms.txt and does not treat it as a ranking signal — John Mueller addressed it publicly in 2025 and compared it to the long-dead meta keywords tag. So the claim that adding llms.txt makes AI engines fetch you first is false.
It is not useless either, but the usefulness is specific. Two groups read it. Developer documentation sites: coding agents follow it to find APIs and tutorials — OpenAI, Anthropic, and Gemini all publish one for their own docs. And platforms that treat it as infrastructure: Mintlify, GitBook, Yoast SEO, AIOSEO, and Wix generate one automatically. Data from Chrome's Lighthouse docs: llms.txt is now part of the Agentic Browsing checks, but the bar is deliberately low — it only flags a server error when fetching the file, and treats a 404 as Not Applicable, so the official stance is still that the file is optional ([Lighthouse docs](https://developer.chrome.com/docs/lighthouse/agentic-browsing/llms-txt)).
Our advice: treat it as a cheap, possibly useful bonus, not a switch. If you run developer or product docs, do it — the payoff is clear. For a pure marketing site it ranks below crawlable body copy, complete structured data, and extractable answer blocks; those three drive AI citations. That is why the single-page check on this site lists llms.txt as a gate rather than a scored category.
Three steps to an llms.txt file
- Enter your site name and a one-line summary — they become the H1 and the blockquote, the first two lines an agent reads.
- Paste your key pages, one per line. A bare URL works; "[Title](URL): note" follows the spec more closely and the note is preserved verbatim.
- Generate, copy, and upload to your site root. Confirm https://your-domain/llms.txt opens directly. Subpaths work too, e.g. /docs/llms.txt.
FAQ: the spec, whether it works, and the limits
How is llms.txt different from sitemap.xml?
sitemap.xml is a complete URL list for crawlers — exhaustive by design. llms.txt is a curated guide for AI — selective by design. They complement each other: the sitemap helps engines discover pages; the llms.txt helps engines understand which pages matter most and how they relate.
Does Google read llms.txt?
No. Google has said it does not use the file and does not treat it as a ranking factor. It is read by other AI agents, developer tooling, and checkers like Chrome Lighthouse. So it is in the no-downside, no-ranking-effect bucket — do not treat it as a switch.
Does the file have to live at the root?
No. The spec allows any subpath, and a file only covers pages under that path: /docs/llms.txt covers /docs/, the root file covers the whole site. When several exist, an agent should use the most specific one.
Does the generator crawl my website?
No. Generation happens entirely in your browser — nothing is crawled or uploaded. You provide the page list, which matches the intent of llms.txt: a hand-curated guide by the site owner, not a mirror of the whole site.
Should I also publish Markdown versions of my pages?
v2 of the proposal recommends it: serve a .md version at the same URL (page.html.md or page.md) and point to it with rel="alternate" type="text/markdown". Then an agent following your llms.txt lands on clean Markdown instead of HTML wrapped in navigation. Optional, but better.
How do I verify it works?
After uploading, open https://your-domain/llms.txt to confirm it is reachable, then run the free Single-Page GEO Check on your site — the llms.txt item will show found and whether it is spec-compliant.
How many links should it contain?
Curated, not exhaustive. 10–30 core pages (home, products, pricing, docs, best articles) work best; hundreds of links dilute the guide's value. Update the file whenever your key pages change.
How does llms.txt relate to robots.txt?
Different jobs. robots.txt says who may crawl; llms.txt says what to read first once they do. Access control versus a reading guide — neither replaces the other, and llms.txt cannot undo a robots.txt block.
Related reading
Haven't checked your page yet? Run the free GEO check first →