Does llms.txt actually work, and is it worth setting up for a client?

Does llms.txt work means whether AI engines actually read the file and use it to inform citations or answers. The short answer: some do, most don't. Whether you set it up for a client depends entirely on which engines matter to their audience.

This guide gives you the adoption table, the setup checklist, and the reasoning behind both.

What llms.txt Is and How It Works

llms.txt is defined as a Markdown file served at /llms.txt that tells AI engines which pages on a site matter most and what they contain. The proposal comes from Jeremy Howard at Answer.AI and lives at llmstxt.org.

The file uses a specific structure: an H1 with the site name (required), an optional blockquote summary, optional paragraphs, then H2 sections with bulleted Markdown links. Each link can include a colon and a short description. An H2 named "Optional" holds links that can be skipped when context windows are tight.

A companion file at /llms-full.txt may hold the full text of linked pages, so an engine can read everything without crawling.

The idea: give AI engines a map instead of making them guess.

Why Adoption Matters More Than the Spec

The file refers to a proposal, not a standard. No governing body enforces it. No engine is required to read it. Adoption is what determines whether the work is worth doing.

Here's what changes that calculation: if the engines your client's audience uses don't read the file, it does nothing. If they do, it gives you a lever to control what gets cited.

For agencies, this means that llms.txt is a targeting decision. You set it up when the client's audience uses engines that read it, and you skip it when they don't.

AI Crawler Adoption Table and Setup Checklist

AI Crawler Adoption Table

This table shows which engines have publicly confirmed reading llms.txt, as of the sources available:

Engine	Reads llms.txt	Public Statement
Google Search	No	John Mueller: not aware of any AI system using it
ChatGPT (OpenAI)	Not confirmed	No public statement
Claude (Anthropic)	Not confirmed	No public statement
Perplexity	Not confirmed	No public statement
Gemini	Not confirmed	No public statement
Bing	Not confirmed	No public statement

Should You Set It Up? Checklist

Use this checklist to decide whether llms.txt is worth the time:

• Does the client's audience use AI engines other than Google Search? If the answer is no, skip it.
• Has any engine the client targets published a statement saying they read llms.txt? If no, the file is speculative.
• Does the client have a small, focused set of pages they want AI engines to prioritize? If yes, the file gives you a way to declare priority — if an engine reads it.
• Is the client's site large enough that crawlers might miss key pages? If yes, and if an engine reads the file, it helps. If no engine reads it, it doesn't.
• Can you set up the file in under an hour? If yes, and the client is curious, it's low-cost to try. If it takes longer, wait for adoption.
• Is the client in a vertical where AI answer engines drive meaningful traffic? If no, the file won't move the needle even if it works.

A workable default if you're unsure: Wait. Monitor public statements from OpenAI, Anthropic, Perplexity, and Google. Add the file when an engine your client's audience uses confirms they read it.

How to Set Up llms.txt (If You Decide To)

If the checklist says yes, here's the process:

1. Create a plain text file named llms.txt in the site root.
2. Add an H1 with the site or project name: # [Client Name]
3. Add an optional blockquote with a one-sentence summary: > [What the site does in one sentence]
4. Add H2 sections for each content category, with bulleted Markdown links below each heading.
5. Use this format for links: - [Page Title](https://example.com/page): Optional short description
6. Add an H2 named "Optional" with links that can be skipped when context is limited.
7. Upload the file to /llms.txt on the domain.
8. Test it by visiting https://[clientdomain.com]/llms.txt in a browser. You should see plain Markdown.

Do not add licensing fields, usage terms, contact info, or metadata. The proposal defines none of those, and adding them breaks the spec.

Example llms.txt File

Here's what a finished file looks like for a fictional agency client:

# [Client Name]

> [One-sentence description of what the site offers]

## Services

- [Service Page Title](https://example.com/services/service-one): What this service does
- [Service Page Title](https://example.com/services/service-two): What this service does

## Case Studies

- [Case Study Title](https://example.com/case-studies/example): What the client achieved

## Optional

- [About Page](https://example.com/about): Team and company background

Replace bracketed placeholders with the client's actual content. Keep descriptions under ten words.

Frequently Asked Questions

Question: Does OpenAI use llms.txt?

OpenAI has not published a statement confirming that ChatGPT reads llms.txt. Until they do, assume it doesn't. Check OpenAI's developer documentation directly before you tell a client it works.

Question: Does Anthropic read llms.txt?

Anthropic has not confirmed that Claude reads the file. The same rule applies: wait for a public statement from Anthropic before you bill for setup.

Question: How does llms.txt work if no one reads it?

It doesn't. The file only works if an AI engine's crawler checks for it and uses the links inside to prioritize content. Without adoption, it's a text file that does nothing.

Question: Is llms.txt worth it for a small client site?

Not yet. Small sites with clear navigation don't need a map file unless an engine confirms they read it. Save the time for work that moves the needle today.

Question: Does llms.txt help with Google Search rankings?

No. It does nothing for traditional Search rankings, AI Overviews, or any other Google product.

Question: Who actually uses llms.txt right now?

The proposal is public, but no major AI engine has confirmed adoption. Individual developers and early adopters use it, but that doesn't mean engines read it. Wait for named adoption.

What to Tell a Client Who Asks About It

When a client asks whether you should set up llms.txt, here's the answer:

Right now, no major AI engine has publicly confirmed they read the file. We can set it up in under an hour if you want to be ready in case adoption happens, but we don't recommend billing for it as a visibility driver until an engine your audience uses confirms they read it.

That keeps the client informed without overpromising.

If the client insists, set it up as a low-cost add-on. Track public statements from OpenAI, Anthropic, Perplexity, and Google. Update the client when adoption changes.

Key Takeaways

• llms.txt is a proposal, not a standard, and no major AI engine has confirmed they read it.
• The file only works if an engine's crawler checks for it and uses the links to prioritize content.
• If your clients' audiences use AI engines that haven't confirmed adoption, a workable default is to wait until an engine they care about confirms they read the file.
• If you set it up, use the exact spec from llmstxt.org: H1, optional blockquote, H2 sections with bulleted links, and an Optional section.
• Monitor public statements from OpenAI, Anthropic, Perplexity, and Google, and update clients when adoption changes.
• Don't bill for it as a visibility driver until you can name an engine that reads it.

See It on a Client's Site

If you'd rather see how AI engines actually read a client's content today — with or without llms.txt — AISO Studio offers a free 7-dimension content audit at aiso.studio/audit. No account needed for the first three audits.

The audit scores content across fact-checking, AEO readiness, WCAG accessibility, readability, SEO, engagement, and GEO. Every factual claim is verified, and sentences that claim evidence without naming a source are flagged.

For agencies running multiple clients, the platform includes white-label PDF reports, client portal pages with your branding, and one-click bundled scans. Try the full platform with a 14-day trial — no credit card required. Cancel any time from the dashboard.

Start at aiso.studio/audit.