llms.txt: What It Is, How to Write One, and Whether It Actually Gets You Cited

llms.txt is a plain Markdown file you publish at your domain root to tell AI engines which pages on your site are worth reading and why. It is not a standard yet, not universally supported, and will not single-handedly get you cited by ChatGPT overnight — but it takes under an hour to create, costs nothing to maintain, and is already confirmed to influence citation behavior on Perplexity and Anthropic's Claude. That risk/reward ratio is hard to argue against.

Here is everything you actually need to know to create one that works.

What llms.txt is — and what it is not

Jeremy Howard, co-founder of Answer.AI, published the original proposal at llmstxt.org in September 2024. The premise is simple: AI answer engines struggle with full websites. They have to parse HTML, navigate JavaScript, skip ads and boilerplate, and guess which of your two hundred pages actually answers the question being asked. llms.txt solves that by giving them a structured shortcut.

Think of it as a curated index — a document you write that says: here is what my site does, here are the pages that matter, here is a one-line explanation of what each page answers.

Three files, three completely different jobs:

robots.txt tells crawlers where they are allowed to go. sitemap.xml lists every destination. llms.txt tells them which stops are actually worth making. You need all three; they do not replace each other.

What the format looks like

The spec calls for plain Markdown with a specific structure. Here is a working skeleton:

> # Your Site or Brand Name > > > One-paragraph summary of what your site does and who it serves. This is what AI engines read to decide if your domain is relevant to a query. > > ## Core Pages > > - Page Title: What this page answers in one direct sentence. > - Service Name: Who this service is for and what problem it solves. > > ## Resources > > - Guide Title: What the reader learns from this guide.

The H1 is your site name. The blockquote is your site summary — write it as if answering "what does this website do?" in one paragraph. H2 sections group related pages. Each bullet is a Markdown link followed by a colon and a short, honest annotation.

What makes a good annotation: state exactly what the page answers, not what it sells. "Step-by-step checklist for setting up Google Business Profile for a service-area business" is useful to an AI engine deciding whether to retrieve your page. "Our comprehensive guide to dominating local search" is not.

One optional companion: if a page is long or complex, you can also publish llms-full.txt with the actual content of your key pages concatenated. This is more effort and only useful for documentation-heavy sites — for most small businesses, a clean llms.txt is sufficient.

Where to host it

One rule: https://yourdomain.com/llms.txt — at the root of your primary domain, not in a subfolder, not on a subdomain.

• WordPress: Drop the file in your site root via FTP or your host's file manager (same directory as wp-config.php).
• Next.js: Place it at /public/llms.txt — Next.js serves everything in /public as static files.
• Webflow / Squarespace / Wix: Use the custom code or file upload feature to place a static file at the root.
• Static site generators: Put it in the build output root (same level as index.html).

Verify by loading the URL in a browser. You should see plain text — not HTML, not a file download prompt. If you see a download dialog, your server is sending the wrong Content-Type header; set it to text/plain.

Does it actually work today?

Honest answer: partially, and unevenly across platforms.

Confirmed support: - Perplexity has publicly confirmed it retrieves llms.txt and uses it to prioritize which pages to read when assembling answers. - Anthropic's Claude (Claude.ai and Claude Desktop retrieval workflows) respects llms.txt directives. - IDE agents — Cursor, Continue, Cline, Aider — actively look for llms.txt when you point them at a documentation site. This is the highest-confidence use case today.

Observable but unconfirmed: - ChatGPT / OpenAI has not officially documented llms.txt support. Practitioners who publish well-formed files report correlated improvements in ChatGPT citation patterns, but this cannot be verified against OpenAI's retrieval internals.

No confirmed support: - Google has not confirmed that Gemini or AI Overviews use llms.txt. Google's preferred signal stack remains robots.txt, sitemap.xml, and structured data.

A SE Ranking analysis of 300,000 domains found roughly 10% adoption as of early 2026 — meaning nine out of ten sites have not published one yet. That is actually a window: in categories where most competitors have not filed llms.txt, your curated file is the only structured signal an AI engine has when comparing sources.

This is a proposed convention gaining real traction, not an official standard. Do not expect it to single-handedly fix weak content or an untrustworthy domain. But as a low-cost, additive signal on top of solid GEO foundations, it earns its hour of setup time.

The actual impact on citation rates

Treat llms.txt as a directional signal, not a citation guarantee. Brands publishing well-curated files report modest but measurable citation rate improvements on Perplexity and Claude — particularly for niche queries where the annotated page is the clearest answer available.

The bigger compounding effect is in developer and agentic contexts: when someone uses an AI coding assistant or research agent and points it at your documentation or service pages, llms.txt is often the first thing the agent reads to understand your site. That is a different kind of citation — not a consumer search result, but a business tool that now treats your content as authoritative.

The bar to publish llms.txt is low enough that the right question is not "should I?" — it is "what is taking me more than an hour?" Write it for the platforms that already support it, file it, and revisit quarterly as the standard matures.