llms.txt: What it is, what Google says and when it's worth having

llms.txt is a markdown text file published at the root of a domain (yourdomain.com/llms.txt) that summarizes what a site is and which pages matter most, written to be read by a language model or an AI agent. Jeremy Howard, co-founder of Answer.AI, proposed it in September 2024, and its specification was revised in August 2026 with version 2.

The question we hear most is not what it is, but whether it helps you show up in ChatGPT or Gemini. With the data available in October 2026, there is no evidence that it does, but getting it right costs almost nothing. This article covers what Google and the specification itself say, which platforms generate it automatically, and how to implement it on WordPress, Shopify or Wix.

What llms.txt is and how it differs from robots.txt and sitemap.xml

When a language model reads a website, it first runs into menus, scripts and styles, and only then reaches the content. llms.txt gives it a clean index: a title, a summary of what the site is, and a list of links with a one-line description each. It doesn’t block or authorize anything. It is a map curated by the site owner.

• robots.txt: read by search and AI crawlers, it tells them which URLs they can or can’t access.
• sitemap.xml: read by search engines, it lists the URLs you want them to discover.
• llms.txt: read by language models and AI agents, it summarizes the site and points to the key content.

There is one prerequisite that doesn’t depend on this file: AI bots must be able to get into your website. If you are blocking them in robots.txt, llms.txt is of no use. We explain how to check this in Is my website blocking AI? and how to set up the file in our robots.txt guide.

What Google says about llms.txt

Google Search’s position has been the same since 2025: it does not use llms.txt and has no plans to. Gary Illyes confirmed it at a Search Central Live event, and John Mueller compared the file to the keywords meta tag. Google’s guide to optimizing for AI features also lists it among the practices you don’t need to appear in AI Overviews or AI Mode (Google’s guide).

We heard it first-hand at Search Central Live Deep Dive Europe, held in Barcelona from September 30 to October 2, 2026 (here is our summary of the event). John Mueller explained that Google doesn’t use it because it doesn’t need it: it already has its own systems to analyze pages and content. What he did not do was rule out that other bots and models, without that infrastructure, do read it, or that Google may adopt it in the future.

In June 2026, on the Search Off the Record podcast, Mueller added a nuance about discovery. Because the site writes the file itself, an AI system can’t rely on it to decide which website is better than another. He did see a narrower role once an agent is already on the site and needs to find its way around, like a store directory for someone who has already walked in (summary at Search Engine Journal).

Google isn’t a single block, either. Since Lighthouse 13.3, released in May 2026, Chrome’s auditing tool includes an agentic browsing category that checks whether a site offers llms.txt. It is an experimental, optional audit: if the file doesn’t exist and the server returns a 404, the check is marked as not applicable, and it is only flagged when there are server errors. Chrome’s documentation describes the file as a machine-readable summary that saves agents crawling time. In other words, Search says it isn’t needed to appear in AI answers, while Chrome considers it useful for agents that browse a website (analysis at Search Engine Journal).

What the data says: does it help you appear in AI answers?

The available studies point in the same direction. SE Ranking analyzed around 300,000 domains and found no relationship between publishing llms.txt and how often a site is cited in AI answers. No major provider has publicly confirmed that it reads the file when crawling third-party websites. OpenAI, Anthropic and Google do publish one for their own developer documentation, and that is where the file has a real use: technical documentation and programming tools. According to the author of the proposal, coding agents use it reliably.

General adoption is low. An analysis of the web’s 100,000 largest domains (SEOmator, September 2026) found llms.txt on 8.6% of them, and about a third of those files had been generated by the platform itself, not by the site owner.

Who has adopted it: platforms and SEO tools

Adoption comes mostly from platforms, not from SEO teams. Shopify has published a default llms.txt on all its stores since 2026, which industry estimates put at around 7 million. Wix generates and maintains it automatically. On WordPress, Yoast SEO, Rank Math and All in One SEO include their own generator, and Yoast also added it to its Shopify app in March 2026. Chrome, for its part, audits it in Lighthouse. These platforms hold a very significant market share, which explains why the file is becoming common even though Google Search doesn’t use it.

Is it worth having? Cost versus benefit

There is an argument in favor that doesn’t depend on anyone reading it today: many practices started as community conventions and search engines adopted them once they became widespread. The clearest case is robots.txt, which began in 1994 as an informal agreement among webmasters and wasn’t standardized until 2022 (RFC 9309). The argument has a counterpoint, because other proposals, such as the keywords meta tag Mueller compares llms.txt to, fell by the wayside. Nobody knows which side this file will end up on. Mueller himself estimated that agentic systems could take anywhere from six months to a year, or longer, to agree on a format.

That gives us a practical criterion: what matters is the effort of creating and maintaining it. If your platform or a plugin generates it in a few minutes, turning it on is a near-zero-cost decision. If it means hand-picking hundreds of URLs and reviewing them every month, that time pays off more in content, structured data or earning quality mentions. And an outdated file can do harm, because an agent that reads it will pick up broken links or obsolete information.

How to create the llms.txt file (specification v2)

Version 2, published on August 10, 2026, keeps the format of the first one. The only required element is an H1 heading with the site name. Everything else is optional, although worth including:

1. An H1 heading with the name of the site or project.
2. A summary in a blockquote (>) explaining what the site is, who it is for and what it offers.
3. One or more paragraphs with additional context.
4. Sections with an H2 heading, each with a list of links in the format - [name](url): short description.
5. A section called Optional with secondary content, which an agent can skip if it is short on context.

An example for a corporate website:

# Company name

> Hydraulic engineering consultancy in Girona. Water supply and sanitation projects for city councils and businesses since 2005.

We work in Catalonia and Aragon. Timelines and prices depend on each project.

## Services
- [Water supply](https://www.example.com/services/water-supply): design and construction management of distribution networks
- [Sanitation](https://www.example.com/services/sanitation): sewers, purification and wastewater treatment

## Company
- [About us](https://www.example.com/company): team, certifications and track record
- [Contact](https://www.example.com/contact): form, phone and address

## Optional
- [Blog](https://www.example.com/blog): technical articles and regulations

What v2 adds is mainly a discovery improvement. The links in the file should point to model-friendly content, such as the markdown version of each page (adding .md to the URL or replacing the extension). An llms.txt inside a subdirectory covers the pages under that path, and the most specific one takes precedence.

You can also declare link relations, in the HTML or in an HTTP header, so that an agent finds the file and the markdown versions of each page without having to guess URLs. The describedby relation points to the llms.txt that covers the page, and the alternate relation with the type text/markdown points to the markdown version of that page. In HTML they are added as link elements in the head. In an HTTP response, the equivalent goes in the Link header.

You can read the details in the specification and in the v2 changelog.

A note on llms-full.txt: it is not part of the specification in any of its versions. Some platforms generate it (Shopify serves it alongside /llms.txt, and technical documentation platforms include it), but a corporate website or a standard ecommerce store doesn’t need it.

How to implement it on WordPress, Shopify and Wix

WordPress

In Rank Math, turn on the LLMS Txt module in the Rank Math SEO dashboard and configure it under General Settings > Edit llms.txt, where you choose the content types and the description of each item. Pages set to noindex are excluded automatically (Rank Math guide).

In Yoast SEO, the option is under Settings > AI tools > llms.txt, with manual selection of the pages you want to include. It is not available on multisite installations (Yoast guide). All in One SEO includes an equivalent generator.

If you don’t use any of the three, the free Website LLMs.txt plugin generates and updates the file automatically and respects the noindex settings of Yoast, Rank Math, SEOPress and All in One SEO. The manual alternative is to create the file and upload it to the domain root via FTP.

Always review the list the plugin generates. By default it tends to include posts and pages, and if your blog has hundreds of old articles, the sensible move is to limit it to services, categories and reference content.

Shopify

Shopify publishes a default llms.txt on every store, available at yourdomain.com/llms.txt. That file mirrors /agents.md, the description Shopify generates for agents. To customize it, edit the theme’s agents.md.liquid template, which affects /agents.md, /llms.txt and /llms-full.txt at once. The llms.txt.liquid template is only needed if you want /llms.txt to differ from the rest, and from that point on its maintenance is manual (Shopify documentation).

You can also manage it with Yoast SEO for Shopify, which has included an llms.txt option with weekly updates since March 2026. On Liquid themes, how much you can customize depends on the variables Shopify exposes, so review what the default file includes before leaving it as it is.

Wix

Wix generates and maintains the file automatically. You can view and edit it in the dashboard, under SEO and GEO > Tools and settings > Go to llms.txt, and switch it off with the opt-out toggle if you don’t want it (Wix help).

Other platforms and custom builds

The usual route is to create the file by hand or with a generator and upload it to the domain root. Firecrawl offers a free tool that generates it from a crawl of the website, although the result needs cleaning up before you publish it.

How to check that it is done right

Before signing off on the file, check these points:

1. It is accessible at the exact path https://yourdomain.com/llms.txt and returns a 200 status code.
2. The markdown is valid. Links follow the [name](url) format, with no space between the bracket and the parenthesis.
3. It has a single H1 and only a few H2 sections.
4. It only includes canonical, indexable URLs that return 200, with no redirects or broken links.
5. The summary is clear and factual, without promotional language. If every site describes itself as the best in its sector, the summary stops adding information.
6. It is short enough to fit in a model’s context, as the specification asks.
7. It is kept up to date. If a plugin generates it, confirm how often it refreshes. If it is manual, review it whenever the site structure changes.

To measure whether anyone reads it, look at your server logs. Google Analytics isn’t suited to this because bots don’t run its tracking code. Look for requests to /llms.txt from user agents such as GPTBot, OAI-SearchBot, ClaudeBot or PerplexityBot. With Chrome 150 or later you can also run Lighthouse’s agentic browsing category from the developer tools.

Our recommendation

On a corporate website or an ecommerce store, turning on llms.txt is a task of a few minutes that commits you to nothing, which is why it is part of the technical setup of our SEO for AI (GEO) service. What we don’t recommend is selling it, or buying it, as an AI visibility strategy.

Whether ChatGPT, Gemini or Perplexity cite a brand depends on the same things as always: useful, well-structured content, authority, and mentions in sources the models consider reliable, all on top of solid SEO. We cover it in our guide to SEO for ChatGPT.