llms.txt: What It Is and Whether It Actually Matters

llms.txt is a proposed convention: a Markdown file, placed at a site’s root or at a path within it, that gives AI agents a short summary of the site and a curated list of links to the content worth reading. It matters for one audience and not for another. For Google Search it does nothing: Google’s documentation says such files are not needed and “won’t negatively or positively impact your visibility or rankings”. For agents that look for the file while helping a user, especially coding assistants reading software documentation, it can shorten the path to the right page. Whether a site should publish one depends on which of those two audiences it actually serves.

What the file is

The convention is described in a public proposal at llmstxt.org, now in a second version. It is not an IETF or W3C standard and no search engine is obliged to read it. The proposal’s format is deliberately simple Markdown in a fixed order:

1. An H1 with the name of the site or project. This is the only required section.
2. A blockquote with a short summary.
3. Optional paragraphs or lists with further context (no headings).
4. Optional sections under H2 headings, each holding a list of links, written as [name](url) with optional notes after a colon.

An H2 section titled “Optional” is used by convention for secondary links an agent can skip when it needs a shorter context.

The proposal adds two companion ideas: publishing clean Markdown versions of important pages at predictable URLs (for example page.html.md), and pointing to those versions and to the covering llms.txt through standard link relations, rel="alternate" type="text/markdown" and rel="describedby". A file at /docs/llms.txt is meant to cover everything under /docs/.

What it is not

The proposal itself separates llms.txt from the files it is sometimes confused with:

llms.txt grants or denies nothing. It cannot block a crawler, opt a site out of AI training, or change how any AI product treats the site. Access and training decisions belong to robots.txt tokens and the controls each provider documents.

Google Search: no effect

Google’s position is stated in two places. Its guide to AI features says site owners do not need to create new machine-readable files, AI text files or special markup to appear in AI Overviews or AI Mode (AI features and your website). Its documentation changelog entry of June 2026 says llms.txt files aren’t needed for Google Search and “won’t negatively or positively impact your visibility or rankings” (Google Search documentation updates).

So the answer to “Is llms.txt a Google ranking signal?” is no, in either direction. Publishing one does not help Google rankings, and not publishing one does not hurt them.

Lighthouse: an audit that is optional for now

Chrome’s Lighthouse added an agentic browsing audit category that includes an llms.txt check. Its documentation, last updated May 5, 2026, describes the behavior (Lighthouse llms.txt audit):

• If the server returns a 404 for /llms.txt, the audit is marked Not Applicable, “as providing the file is optional at the moment”.
• Lighthouse flags the page if a server error occurs when it tries to retrieve the file.

That second point is the one to act on. A site without llms.txt is fine in Lighthouse. A site whose /llms.txt path returns a 500 because of a misconfigured route or a broken plugin is flagged. If a CMS plugin generates the file, confirm the URL returns either the file with a 200 status or a clean 404.

Lighthouse is a diagnostic tool. Its audit result does not feed Google Search rankings, and an N/A or a pass in this category says nothing about visibility in Search.

Where llms.txt can actually help

The proposal states its expected use: agents read the file on demand while assisting a user, and follow links to the detail they need. The clearest fit is documentation:

• Developer documentation. A coding assistant asked how to call an API can read a short index and jump to the right reference page instead of parsing navigation-heavy HTML.
• Policy-heavy sites. A business whose pricing, returns or eligibility rules are spread over many pages can point an agent to the authoritative ones.
• Sites with Markdown sources. Documentation generators that already produce Markdown can publish the .md versions the proposal describes at little extra cost.

Neither the proposal nor the provider documentation discussed here reports a measured effect on visibility or citations in any AI product. OpenAI’s crawler documentation uses llms.txt only to point readers to its own documentation index, and Anthropic’s crawler help page does not mention the file; neither describes it as a factor in what their search features show. Google says it has no effect on Search. Treat the file as a convenience for agents that choose to read it.

A decision framework

Publish an llms.txt file when most of these are true:

• the site has content that agents are asked about, such as documentation, product specifications or policies;
• the content is spread across many pages and a curated reading order would help;
• someone will keep the file current when pages move;
• the links in it can point to clean, stable URLs.

Skip it, or postpone it, when:

• the goal is Google rankings or AI Overview visibility (no effect on either);
• nobody owns its maintenance (a stale reading list sends agents to outdated or missing pages);
• the site’s important information is already on a few clear pages.

Either way, make sure /llms.txt does not return a server error.

An illustrative file

The example below follows the proposal’s structure for a hypothetical accounting software company. The URLs are placeholders.

# Ledgerly

> Ledgerly is cloud accounting software for small businesses in the United States. This file lists the pages that explain pricing, data handling and the public API.

Prices are listed in US dollars and change with plan revisions; the pricing page is the authoritative source.

## Product
- [Plans and pricing](https://example.com/pricing.md): Current plans, limits and billing terms
- [Data retention policy](https://example.com/legal/data-retention.md): How long records are kept after cancellation

## API
- [API quick start](https://example.com/docs/api/quickstart.md): Authentication and first request
- [Invoices endpoint](https://example.com/docs/api/invoices.md): Fields, errors and rate limits

## Optional
- [Release notes](https://example.com/changelog.md): Version history

Expected result: an agent that fetches the file finds the summary, sees which pages are authoritative and can skip the “Optional” section when context is tight. Verification: request /llms.txt and each linked URL and confirm a 200 status; ask an AI assistant a question about the product while giving it only this file as a starting point, as the proposal suggests. Limitation: none of this guarantees that any AI product will fetch the file.

Maintenance rules

• Regenerate or review the file when linked pages move; a 404 inside llms.txt sends agents to a dead end.
• Keep the summary factual. The file is read as a statement of what the site says about itself.
• Do not list private, gated or unpublished URLs; the file is public.
• If a plugin generates the file automatically, review its output once, since generated lists can include every post rather than a curated set.

The short answer

llms.txt is an optional, agent-facing reading list. It does not affect Google rankings, Lighthouse treats its absence as not applicable, and its value depends on whether agents that help your users actually read it. Publish one if you have documentation or policy content worth guiding agents to and someone to maintain it; skip it if the goal is search rankings.