Free llms.txt generator
Fill in your site name, a one-line summary and the URLs that matter. You get a spec-correct llms.txt you can copy or download and drop at the root of your domain. Nothing is uploaded — the file is built in your browser.
Honest framing up front: no major AI engine guarantees it reads llms.txt today. It is a five-minute, low-cost hygiene file that agents and developer tools do use — not a ranking trick. More on that below.
llms.txt generator
Your site
Becomes the H1 — the only line the spec requires.
Used to turn relative paths like /pricing into absolute URLs.
Becomes the > blockquote. Say what you are, who it is for, and what makes it different — in one sentence.
Free prose after the blockquote. Blank lines start a new paragraph. Headings are not allowed here — put those in the link sections below.
Link sections
Each section becomes an ## heading with a Markdown list under it. Keep it to the pages that genuinely answer questions.
The spec gives Optional a special meaning: URLs a model can skip when it needs a shorter context. Put secondary pages here.
llms.txt preview
0 sections · 0 links · 17 bytes
# Your Site Name
Save it as llms.txt and serve it at https://yoursite.com/llms.txt
Suggestions
- · Add your site or product name — it becomes the H1, the only required line in the spec.
- · Add the one-line summary. It becomes the blockquote, and it is the first thing a model reads.
- · Add at least one link — a title and a URL — so the file is actually useful.
- · Section “Docs” has no links yet, so it is left out of the file.
What llms.txt is — and what it is not
llms.txt is a proposal from Jeremy Howard, published in September 2024 and documented at llmstxt.org. The idea is simple: a website is written for humans, full of navigation, scripts and boilerplate, and a language model working inside a limited context window has to wade through all of it. So you publish one small Markdown file at the root of your domain that says, in plain language, what this site is and which handful of pages actually matter.
It is a curation format, not an access-control format. It grants nobody permission and blocks nobody. It also carries no guarantee of being read. Google has stated that no Search system consumes llms.txt, and OpenAI's crawler documentation governs its bots through robots.txt without mentioning the file at all. Independent crawl studies published through 2025 and 2026 found that the overwhelming majority of published llms.txt files were never fetched by anything.
Where it does get used is agent tooling. Anthropic recommends the file in its guidance on writing for agents, OpenAI ships one alongside its Agents SDK documentation, Chrome's Lighthouse added an agentic-browsing audit that checks for it, and coding assistants such as Cursor, GitHub Copilot, Windsurf, Cline and Aider read it to find current docs instead of hallucinating an old API. If you sell to developers, or if anything you publish gets consumed by an agent, that is a real audience. Publishing the file costs about ten minutes. Just be honest with yourself about why you are doing it: it is housekeeping, not a growth lever.
The format, line by line
The spec is deliberately tiny, and the order is fixed. An H1 with the site or project name comes first and is the only required element. Then a blockquote — a line starting with > — holding a short summary with the key facts needed to understand everything that follows. After that you may add any amount of free prose: paragraphs, lists, anything except headings. Then come zero or more H2 sections, each containing a Markdown list of links in the form - [Title](url): note, where the note after the colon is optional but very much worth writing.
One H2 name is special. A section literally titled ## Optional tells a consumer that those URLs can be dropped when a shorter context is needed. Everything that is nice-to-have — changelog, about page, press kit — belongs there, and it should be the last section in the file. This generator enforces that placement for you.
Some sites additionally publish llms-full.txt, which inlines the full Markdown text of the linked pages into a single document. That is a separate, larger artifact — useful for docs sites, overkill for most marketing sites, and not part of the core spec.
Where the file goes
Put it at the root: https://yoursite.com/llms.txt. On Next.js or Vite that means dropping the file into your public/ directory; on most static hosts it goes in the published output folder. Serve it as UTF-8 with a text/plain or text/markdown content type, make sure it returns a 200 rather than a redirect or your SPA's HTML fallback, and confirm your robots.txt does not block it. Open the URL in a private window once you have deployed. A surprising number of llms.txt files quietly return the site's 404 page.
Common mistakes
- Dumping your whole sitemap into it. Four hundred links is a sitemap, not a summary. Ten to forty curated links is the useful range.
- Skipping the notes. The sentence after the colon is where you disambiguate — it is what stops a model guessing which of your three pricing pages is canonical.
- Relative URLs. A consumer may read the file with no memory of where it came from. Always emit absolute https URLs; this tool resolves relative paths for you if you fill in your site URL.
- Letting it rot. A file listing last year's pricing page is worse than no file. Regenerate it when your docs or pricing change.
- Marketing prose instead of facts. Write the pricing, the integrations, the limits. Adjectives do not survive summarisation; numbers do.
- Treating it as a substitute for crawlability. If GPTBot and the other AI crawlers are blocked, or your content only exists after JavaScript runs, llms.txt fixes nothing.
llms.txt vs robots.txt vs sitemap.xml
| File | Answers | Read by |
|---|---|---|
| robots.txt | May you fetch this? | Every serious crawler, including GPTBot, ClaudeBot, PerplexityBot |
| sitemap.xml | What exists here? | Search engines, for discovery and freshness |
| llms.txt | What matters here, and why? | Agents and dev tools today; no engine guarantees it |
They are complements, not alternatives, and the order of effort should follow that table. Get crawler access right first, keep a clean sitemap second, then add llms.txt. If you want the deeper version of this, the complete llms.txt guide for B2B SaaS walks through a full implementation, and generative engine optimization covers the work that actually moves AI answers: being cited by sources the models already trust.
Frequently asked questions
What is llms.txt?
llms.txt is a plain-text Markdown file at the root of your site (/llms.txt) that gives large language models a short, curated map of your most important pages. It starts with an H1 site name, a one-line blockquote summary, optional prose, and H2 sections of Markdown links. It was proposed by Jeremy Howard in September 2024 and is documented at llmstxt.org.
Does llms.txt actually improve my AI search visibility?
There is no evidence that it does. Google has said no Search system reads or acts on llms.txt, and OpenAI's crawler documentation only references robots.txt. Where the file demonstrably gets used is agent and developer tooling — Anthropic recommends it in its guidance for writing for agents, OpenAI ships one for its Agents SDK docs, and coding assistants like Cursor, GitHub Copilot, Windsurf and Cline read it. Treat llms.txt as cheap, useful housekeeping, not as a ranking lever.
Where do I put the llms.txt file?
At the root of your domain, so it resolves at https://yoursite.com/llms.txt. Serve it as UTF-8 plain text (text/plain or text/markdown), return a 200 status, and make sure robots.txt does not block it. Some sites also publish an llms-full.txt containing the full text of the linked pages.
How is llms.txt different from robots.txt and sitemap.xml?
robots.txt is permission — it tells crawlers such as GPTBot, ClaudeBot and PerplexityBot what they may fetch. sitemap.xml is inventory — every URL you want indexed, in machine-readable XML. llms.txt is curation — a short, human-written shortlist of the pages that matter most, with a sentence explaining each one. They solve different problems and you should have all three.
How long should llms.txt be?
Short. The point of the file is to fit comfortably in a context window, so aim for the 10 to 40 links that genuinely answer a buyer's or an agent's questions: docs, pricing, product, security, and support. Push everything lower-priority into the ## Optional section, which the spec says can be skipped when a shorter context is needed.
Keep reading
Publishing the file is the easy part
The harder question is what ChatGPT, Perplexity, Gemini and Claude actually say when a buyer asks about your category — and whether they hand that buyer to a competitor. Perciva monitors those answers and shows you the verbatim response. No email needed to use this generator; the snapshot is separate and also free.