What Is llms txt
An llms.txt file is a short Markdown document placed at a site's root (https://yoursite.com/llms.txt) that tells language models what a site is and which pages matter, without making them guess from a full HTML crawl. It has a required H1 title, an optional summary, and sections of curated links. Two cautions belong in the first paragraph: llms.txt is a proposal described at llmstxt.org, not a finished, universally adopted standard — and publishing one does not guarantee AI citations, rankings, or traffic.
At a glance
- An llms.txt file is a proposed Markdown site map: H1 title, optional blockquote summary, optional detail text, then ## sections of links.
- It is a proposal, not a ratified standard. Tools may read it, ignore it, or implement it differently — do not build anything that depends on it being honoured.
- Link format per section: - [name](url): notes — a name, a URL, and a short note on what the page is.
- It lives at the site root: https://yoursite.com/llms.txt (for many static or Next.js sites, that means the public/ folder).
- Clioos offers a free generator and checker with no account: generate from a site name, summary, and links, then check structure or fetch a public URL.
- Publishing the file is a helpful signal and documentation aid, not a ranking promise. Anyone selling guaranteed AI citations from llms.txt is overselling.
Steps
- Read the proposal first: llmstxt.org describes the file's intended shape (H1 title, optional summary, ## sections of links, optional ## Optional section, and an optional fuller llms-full.txt companion). Clioos is not affiliated with llmstxt.org; its tool follows that convention.
- Open https://clioos.com/llms-txt/ . The generator is free and needs no account. Enter the site name, an optional summary and details, the homepage URL, and the important links you actually want models to find first.
- Generate and copy or download the .txt. Generation is pure templating — no LLM call, no whole-site crawl, no payment — so what you get is exactly the structure you described, in spec order.
- Check the result with the same tool: paste the body, or point the checker at a public URL. It reports whether the body looks like llms.txt, whether an H1 is present, how many sections it has, and any structural warnings.
- Place llms.txt at the site root and confirm https://yoursite.com/llms.txt returns the file. Then stop: do not expect citations, rankings, or traffic from this step alone.
A proposal, not a finished standard
The distinction matters for planning. A finished standard (like robots.txt, with decades of consistent crawler behaviour) can be relied on. A proposal can be read by some tools, ignored by others, and revised without notice. The Clioos tool page states this plainly: llms.txt is “a proposed site root file (see llmstxt.org)” and the tool “follows the llmstxt.org convention.” Nothing on this page should be read as “all AI engines do X with llms.txt,” because no such universal behaviour exists to cite.
That is also why the file is cheap to adopt and cheap to abandon. It costs one small Markdown file at the root and a few minutes of curation. If tools stop reading it, you have lost almost nothing. If they read it, they find your chosen pages first instead of guessing. Treat it as documentation with a possible upside, not as infrastructure.
Worked example (tested 2 November 2026)
Test method, disclosed honestly: this run hit the live llms-txt service's own API on the server (POST /llms-txt/api/generate and POST /llms-txt/api/check on the local service port, 2026-11-02), with a fictional example site — “Example Bakery,” a small bakery site with menus and ordering pages. No account, no payment, and no LLM call were involved: generation is pure templating, and the no-code version of the same job is the form at https://clioos.com/llms-txt/ . The generated file is reproduced verbatim in the companion example guide.
| Stage | Result |
|---|---|
| Generate call | HTTP 200 in 0.03 seconds; site name “Example Bakery”, summary, homepage URL, and 2 important links (Menu, Order) |
| Generated file | H1 “# Example Bakery”, blockquote summary, one ## Docs/Pages section with 3 links (Homepage auto-added plus Menu and Order with notes) |
| Generate-time validation | valid: true, has_h1: true, h1_count: 1, section_count: 1, link_count: 3, errors: [], warnings: [] |
| Check of the generated file | looks_like_llms_txt: true, has_h1: true, section_count: 1, link_count: 3, valid: true, warnings: [], errors: [] |
| Check of plain text without an H1 | valid: false, looks_like_llms_txt: false, with the warning and error “Missing required H1 title (for example, '# My Site').” |
The contrast is the lesson: the checker is structural, not semantic. It confirms the file has the required shape; it cannot confirm the summary is accurate, the links are the right ones, or that any AI tool will ever read the file.
What it can't do
- It cannot guarantee citations, rankings, or traffic: the tool page and its FAQ say this explicitly. Publishing llms.txt is a signal, not a ranking lever.
- It cannot force any tool to read it: as a proposal, adoption varies by tool and may change. Do not gate launches, launches copy, or client promises on it.
- It cannot replace real documentation: a link list with one-line notes is a map, not the territory. The pages it points to still have to be good.
- The generator cannot crawl your site: it templates what you type. If you forget a key page, the file forgets it too — curation is the work, and it is yours.
- The checker cannot judge quality: it reports structure (H1 present, section count, warnings). A structurally valid file can still point at dead links or describe the site badly.
- It is not the official llmstxt.org tool: the Clioos generator is an independent free tool that follows the convention and links to the proposal. Do not cite Clioos output as the specification itself.
Where the file lives
At the site root: https://yoursite.com/llms.txt. On many static sites and Next.js apps that means dropping llms.txt into the public/ folder (or its equivalent) so the URL returns the file as text. If the URL 404s, redirects to the homepage, or returns HTML, the file is not published — regardless of how good its contents are. After placing it, fetch the URL yourself and confirm you see Markdown, not a homepage.
An optional companion, llms-full.txt, can carry a fuller outline for tools that want more detail. It is optional in the same sense as everything else here: useful if a tool asks for it, harmless if none does.
Frequently asked questions
What is an llms.txt file?
A proposed site-root Markdown file that gives language models a short, structured map of a site: title, optional summary, and curated links — so tools can discover important pages without guessing from a full HTML crawl.
Is llms.txt an official standard?
No. It is a proposal described at llmstxt.org. Tools may read it, ignore it, or implement it differently, so do not build anything that depends on it being honoured.
Does publishing llms.txt guarantee AI citations?
No. It does not guarantee citations, rankings, or traffic. Treat it as a helpful signal and documentation aid, not a ranking promise.
Do I need an account or payment for the Clioos generator?
No. Generate and check use is free and does not require an account. There is no subscription or check pack for this tool.
Where should I put the file?
At the site root, so https://yoursite.com/llms.txt returns the file. On many static or Next.js apps that means the public/ folder or its equivalent.
Related guides: llms txt Example · llms.txt Generator & Checker tool · The llms.txt proposal (llmstxt.org) · All guides · Tool: PDF Probe