---
title: "llms.txt"
description: "llms.txt: where the proposal comes from, the file format, Google's official position and the precautions to take before publishing one on WordPress."
language: "en"
locale: "en-US"
type: "definition"
canonical: "https://www.wp-masquerade.com/glossary/llms-txt/"
published: "2026-10-28T08:00:00+00:00"
modified: "2026-10-28T20:07:29+00:00"
translations:
  en-US: "https://www.wp-masquerade.com/glossary/llms-txt/index.md"
  fr-FR: "https://www.wp-masquerade.com/fr/definitions/llms-txt/index.md"
  es-ES: "https://www.wp-masquerade.com/es/glosario/llms-txt/index.md"
  de-DE: "https://www.wp-masquerade.com/de/glossar/llms-txt/index.md"
  it-IT: "https://www.wp-masquerade.com/it/glossario/llms-txt/index.md"
  pt-BR: "https://www.wp-masquerade.com/pt/glossario/llms-txt/index.md"
  nl-NL: "https://www.wp-masquerade.com/nl/woordenlijst/llms-txt/index.md"
---

Language models struggle to read full web pages cluttered with menus, scripts and ads, and their context window is limited. The idea behind llms.txt is to give them a clean, concise table of contents for the site at a well-known address. The file has often been billed as “robots.txt for AI”, a misleading comparison: it grants no permission and forbids nothing.

## Origin and format

The proposal was published on September 3, 2024 by Jeremy Howard on llmstxt.org. The file, placed at the root (`/llms.txt`), is written in Markdown in a fixed order: a level-1 heading with the site name, the only required part; a summary as a blockquote; optional paragraphs; then level-2 sections containing lists of links, each of which can be followed by a short note. A section titled “Optional” groups secondary links that an agent may skip.

```
# Example Pottery Studio
> Handmade ceramics shop in Portland: catalog, workshops and care tips.
## Guides
- [Caring for stoneware](https://example.com/guides/stoneware-care/): washing, oven, dishwasher
- [Choosing your dinnerware](https://example.com/guides/choosing/): materials and uses
## Optional
- [Our story](https://example.com/about/)
```

The proposal also suggests publishing a Markdown version of each useful page, at the same URL followed by `.md`. The site describes itself as a proposal open to contributions: it is not a standard published by a standards body.

## Who actually uses it

As far as Google is concerned, the answer is documented. Its guide to optimizing for generative AI, published on May 15, 2026, states that no special text file for AI is needed to appear in Search. On June 15, 2026, Google added a clarification: creating and maintaining an llms.txt file neither hurts nor helps visibility or ranking in Google Search, and you can keep one for other services that use it.

For other assistants, nothing comparable has been published: their pages about their crawlers describe how they respect [robots.txt](https://www.wp-masquerade.com/glossary/robots-txt/), not how they read llms.txt. The file has mostly been adopted by technical documentation sites, often generated automatically by their documentation platforms. For where this file fits in a visibility strategy, see our definition of [GEO](https://www.wp-masquerade.com/glossary/generative-engine-optimization/).

## For WordPress: should you create one?

The cost is low, the proven benefit is zero for Google and poorly documented elsewhere. If you publish one, by hand or with an SEO plugin that generates it, treat it like any other public file:

- Only list pages that are already public and indexable, as you would in an [XML sitemap](https://www.wp-masquerade.com/glossary/xml-sitemap/): never a staging site, a customer area or an internal document.
- Don't mention the WordPress version, the installed plugins or the login URL: that would help attack bots with [fingerprinting](https://www.wp-masquerade.com/glossary/fingerprinting/).
- If you publish `.md` versions of your pages, make sure they expose neither drafts nor members-only content.
- Check what an automatic plugin generates, and give it the same scrutiny as any other: updates, active developer.

The status of the proposal, its key dates and the precautions to take are covered in our article on [robots.txt, llms.txt and structured data on a secure WordPress site](https://www.wp-masquerade.com/geo-secure-wordpress-llms-txt/). To decide which AI crawlers to let in, also read [our analysis of GPTBot, ClaudeBot and PerplexityBot](https://www.wp-masquerade.com/ai-crawlers-gptbot-claudebot-block/).

## Related terms

- [GEO (Generative Engine Optimization)](https://www.wp-masquerade.com/glossary/generative-engine-optimization/)
- [Robots.txt](https://www.wp-masquerade.com/glossary/robots-txt/)
- [XML sitemap](https://www.wp-masquerade.com/glossary/xml-sitemap/)
- [Fingerprinting](https://www.wp-masquerade.com/glossary/fingerprinting/)
- [Structured data](https://www.wp-masquerade.com/glossary/structured-data/)


## Sources

- [llmstxt.org : The /llms.txt file](https://llmstxt.org/)
- [Google Search Central : optimiser votre site pour les fonctions d’IA générative de la recherche Google](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)
- [Google Search Central : journal des mises à jour de la documentation (15 juin 2026)](https://developers.google.com/search/updates)