llms.txt on WordPress: what it is, who reads it and whether your site needs one

llms.txt is a proposed text file, written in markdown and placed at /llms.txt, that lists a site's most important pages for AI agents. It is a proposal published on llmstxt.org and not a standard, and WordPress does not make one. An SEO plugin may. All in One SEO makes one by default, and Yoast SEO, Rank Math and SEOPress PRO make one once the feature is switched on.

Whether anything reads the file is a separate question, and this page answers it from each company's own documentation. Google says Search ignores it. The crawler pages of OpenAI, Anthropic and Perplexity, and Bing's Webmaster Guidelines, do not say their crawlers read one. The file costs little, and nothing an engine has published says it changes whether your pages are quoted. So find out whether your site already has one before anyone sells you one.

What llms.txt is

The proposal lives at llmstxt.org. That page names Jeremy Howard as its author, and gives September 3, 2024 as the day it was published and August 10, 2026 as the day it was last modified. The current text is version 2, which the proposal's page of changes dates to August 2026. It describes itself in one line: "A proposal to standardise on using an /llms.txt file to provide information to help agents use a website."

The file is ordinary markdown in a fixed order:

A section named "Optional" is, by the proposal's convention, for links that can be skipped. The file sits at the root of the site, or in a folder, where it covers the pages under that folder.

The idea, as the proposal puts it, is that an agent reads the short file first and then follows only the links it needs. The proposal says such files are used most heavily for software documentation, where coding agents follow them to find references and tutorials. Version 2 also proposes a markdown copy of each page at the same address with .md added, and two link relations that point to those copies and to the file.

It is a proposal, not a standard

The proposal's own page calls it a proposal throughout. It closes by saying the specification is open for community input, and that a GitHub repository hosts what it calls an informal overview. It names no standards body. The plugins that generate the file say the same in their own documentation. Rank Math's reads: "The llms.txt is currently a proposal and not an official standard." The changelog of the SEOPress release that added the feature carries a disclaimer that the file is not yet officially supported by major search engines or AI platforms.

What it is not

The proposal draws the same lines itself. It says robots.txt tells automated tools what access is acceptable, while llms.txt is read on demand when an agent needs information, and that a sitemap lists all of a site's indexable pages, which is too much to stand in for a short list. Nothing you write in llms.txt keeps a crawler out or lets one in. Where robots.txt is in WordPress and the sitemap at wp-sitemap.xml cover the other two files.

Who reads it, according to each company

Each row is what that company's own page said when it was read on October 8, 2026. A page that does not mention the file is reported as exactly that. Silence is not a statement that a crawler ignores the file, and it is not support either.

Google is the only one of the five that states a position. Its guide lists llms.txt among things a site owner can ignore for Search, and the documentation updates log records a note added on June 15, 2026 to make that plain. The guide says it is fine to keep such a file for other services that use one, and then: "Doing so will neither harm nor help your site's visibility or rankings in Google Search, as Google Search ignores them." It adds that Google may crawl and index many kinds of files, and that finding the file does not mean it is treated in a special way.

Two other parts of Google do touch the file, and neither is Search. Chrome's Lighthouse has an llms.txt audit, whose documentation, last updated May 5, 2026, calls the file an emerging convention, marks a site with no file as not applicable because providing one is optional, and flags only a server error at that address. And the Gemini API's documentation publishes an llms.txt of its own.

OpenAI, Anthropic and Perplexity publish the file for their own documentation, which is not the same as their crawlers reading yours. On OpenAI's and Perplexity's crawler pages, the only mention of llms.txt is a pointer to the index of that company's own developer documentation. Anthropic publishes one for its developer documentation and one for its help center. Each of those files is an index of the company's own pages. None of the crawler pages says that OAI-SearchBot, GPTBot, ClaudeBot, Claude-SearchBot or PerplexityBot fetches /llms.txt from the sites it visits, or does anything with one.

That an assistant "supports" llms.txt is said widely. On the pages above it is documented by nobody.

What was studied

One study with data is worth knowing, with its limits.

SE Ranking, a company that sells SEO software including tools that track AI search, published it on November 7, 2025. It reports on nearly 300,000 domains. Of those, 10.13% had the file. Its authors found no correlation between having the file and how often a domain was cited in the AI answers they sampled, and say that taking the file out of their prediction model made the model more accurate.

The article does not name the assistants sampled or the dates of the sample, and its own disclaimer says the findings depend on the model and dataset tested. It is a finding of no relation in one dataset, not proof that the file can never matter. It has not been reproduced here.

How a WordPress site comes to have one

WordPress does not make an llms.txt. On a fresh WordPress with no plugin active and any permalink structure other than Plain, the address answers with the site's "Page not found" page and a 404. The check further down covers Plain permalinks.

Four SEO plugins can make one. Each line below is from that plugin's own documentation or source.

So a site that runs All in One SEO has the file unless someone switched it off, and a site that runs one of the others has it if someone switched the feature on. Installing or updating an SEO plugin can therefore be the whole of what an "llms.txt setup" amounts to.

What goes into a generated file is the plugin's choice. Yoast's specification says it includes the five latest updated posts, pages and custom post types, cornerstone content first, and that pages can be picked by hand instead. Rank Math's documentation says you choose the post types and it lists up to 100 items by default, leaving out posts set to noindex.

Check whether your site has one

Make one without a plugin

A static file is enough. Write it in a plain text editor, save it as llms.txt in lowercase, and upload it to the site's root.

# Example Plumbing Co.

> Example Plumbing Co. repairs and installs plumbing for homes and small businesses. This file lists the pages that say what we do and how to reach us.

Opening hours and the area we cover are on the contact page.

## Services

- [Emergency repairs](https://example.com/services/emergency-repairs/): What counts as an emergency and how to call us out
- [Water heater installation](https://example.com/services/water-heater-installation/): What an installation includes

## Company

- [About us](https://example.com/about/): Who we are and the area we cover
- [Contact](https://example.com/contact/): Phone, email and opening hours

## Optional

- [Blog](https://example.com/blog/): Seasonal advice for homeowners

A few things to know about a file made this way:

• The server sends it as plain text. An Apache server with nothing configured for the file answers 200 text/plain, taking the type from the .txt ending.
• It is served whatever the permalinks are. The server finds the file before WordPress is asked, in the same way as a real robots.txt.
• It is yours to keep true. Nothing updates it when a page is renamed or removed.
• Switch a plugin's version off first. Yoast's specification says it will not overwrite an llms.txt that already exists and shows a warning instead, and tells you to disable its feature before creating your own.
• The links are ordinary pages. Version 2 of the proposal prefers links to markdown copies of pages. A WordPress site has none unless a plugin makes them, and Yoast's and Rank Math's documentation describe their files as lists of each page's ordinary address.

The file is public. Anyone can open it, so list nothing in it that you would not link from your menu.

Remove one a plugin made

Switch the feature off in the plugin that made it. These are the places each plugin's documentation gives.

Then run the check from the section above again and expect a 404, or the home page while permalinks are Plain. If the address still answers, look in the site's root for a file named llms.txt and delete it there: a real file is served whatever a plugin's settings say. If a cache or a CDN stands in front of the site, clear it before deciding the file is still there.

Whether to bother

For a site that sells a service or runs a store, plainly:

• It costs little. A short file, or a switch in a plugin you already have.
• Nothing an engine has published says it changes whether you are quoted. Google says Search ignores it. The other four companies' pages say nothing about reading it. The one study with data found no relation.
• It is not access control. It does not block a crawler, allow one, or opt a site out of training.
• Its documented use is narrow. The proposal itself says the files are used most heavily for software documentation read by coding agents. A site with documentation for developers is the case it was written for.

So it is optional and low on the list. Leaving a plugin's file in place is harmless as long as what it lists is accurate. Paying for one as a way into AI answers is paying for something no engine has said it uses.

What does have documented standing is elsewhere. Whether an assistant's crawler can reach your pages at all is decided by robots.txt and by any CDN or host in front of the site, and blocking or allowing AI crawlers on WordPress covers that. What each engine says it looks for, and what is only guessed, is set out in generative engine optimization: what is documented, what was studied and what is guesswork.

This site serves an llms.txt of its own, at /llms.txt.

When to get help

• The address answers with something you cannot trace to a file or a plugin, or a file keeps coming back after you delete it. A one-time fix from WP Ministry covers one issue on one site and starts with a free diagnosis, which gives you a written cause and a fixed quote.
• You want the documented things checked for you: crawler access, indexing and what each engine's own reports show. That is the work described under AI search optimization.