Technical guide
What is llms.txt, and is it worth publishing?
llms.txt is a Markdown file at your site root (/llms.txt) that summarizes your site for language models and AI agents, with links to your key pages. Jeremy Howard proposed it in 2024 at llmstxt.org. It is cheap to create and useful for agents, but as of November 2026 no major assistant has confirmed it affects their answers. GeoRush checks for it and drafts one.
Updated on
What it is, exactly
llms.txt is an open proposal, not an official standard from any standards body. Jeremy Howard (Answer.AI) published it on September 3, 2024, and version 2 reflects two years of real-world use. The idea: web pages are built for people, with menus, ads and JavaScript that make it hard for a model to extract what matters, and a whole website does not fit in a model’s context. A short Markdown file explaining what the site is and where the key information lives saves agents work.
It does not replace robots.txt or your sitemap. robots.txt says what bots may crawl; the sitemap lists every indexable page; llms.txt is a curated, annotated selection of the most useful content, meant to be used at answer time (inference) rather than for training.
The format, section by section
According to the llmstxt.org specification, the file contains these sections in this order:
| Section | Required? | Content |
|---|---|---|
H1 title (# Name) | Yes, the only required one | The name of the project or site. |
Blockquote (> summary) | No | A short summary with the key information needed to understand the rest. |
| Paragraphs or lists | No | Details and how to interpret the links. Any Markdown except headings. |
| H2 sections with link lists | No | Each item: [name](url), optionally followed by : and a note. |
## Optional section | No | By convention, secondary links an agent can skip when it needs a shorter context. |
The proposal also suggests offering clean Markdown versions of important pages (for example page.html.md or page.md) and linking them with rel="alternate" type="text/markdown". The file can sit at the root or under a subpath (/docs/llms.txt), covering the URLs under that path.
An example llms.txt for a business
An example for a fictional software company. Copy it and replace the details with yours. Publish it as plain text at https://your-domain.com/llms.txt.
# Acme Invoicing
> Acme Invoicing is online invoicing software for freelancers and small businesses in the US and Canada. Plans from $0 per month; 14-day free trial.
Key facts: founded in 2019. Integrates with major banks and online store platforms. Support in English and French.
## Product
- [Features](https://your-domain.com/features.md): what the software does, module by module
- [Pricing](https://your-domain.com/pricing.md): plans, limits and what each one includes
- [Integrations](https://your-domain.com/integrations.md): full list and requirements
## Help
- [FAQ](https://your-domain.com/faq.md): e-invoicing, taxes and accounts
- [Comparison](https://your-domain.com/compare.md): how we differ from other options
## Optional
- [Blog](https://your-domain.com/blog.md): articles and updates
Actual adoption status (November 2026)
What can be checked in public sources, separated from what cannot:
| Claim | Status |
|---|---|
| Thousands of sites publish llms.txt | The proposal itself (v2) says so. It is common mainly in technical documentation; several documentation platforms generate it automatically. |
| OpenAI, Anthropic and Google publish llms.txt | Yes, for their developer documentation. That does not mean their assistants use it to decide what to recommend. |
| Lighthouse checks it | Chrome Lighthouse has an llms.txt audit among its agentic browsing checks: it flags a server error when fetching the file; a missing file (404) is marked “not applicable”, because the file is optional. |
| Coding agents read it | This is the most widespread use, according to the proposal: agents that consult documentation to write code. |
| ChatGPT, Gemini, Claude or Perplexity use it to choose sources | No official confirmation as of November 2026. |
| It helps you appear in Google AI Overviews | No. Google explicitly says you do not need AI text files to appear in AI Overviews or AI Mode. |
How to create yours in 20 minutes
- 1
Write the summary
One or two sentences: what you do, for whom, where, and one differentiating fact. Use the same terms as on your site and profiles.
- 2
Pick 5 to 15 links
The pages that answer what customers ask most: features, pricing, comparisons, FAQ, contact.
- 3
Add a note to each link
One sentence saying what the agent will find there. It helps decide whether the page is worth opening.
- 4
Publish and link
Upload it to the root as plain text, check it returns a 200 status and, if you can, offer .md versions of the linked pages.
- 5
Keep it current
Review the file whenever prices, plans or key pages change. An outdated llms.txt spreads wrong facts.
How GeoRush helps
The AI technical section of the SEO and technical audit checks whether you have llms.txt and offers a generator that drafts one from your site, alongside the robots.txt and JSON-LD generators. The free check also reviews it as part of your AI-readiness grade. And because llms.txt is only one piece, AI visibility measures whether your changes actually move your mentions.
Is your site ready to be read by AI?
The free check reviews robots.txt, llms.txt, indexing and structured data, and gives you an A–E grade.
Sources
Official documentation and primary sources checked on November 16, 2026. These policies change often: check the source before making important decisions.
Frequently asked questions
Keep reading
robots.txt for AI bots
What each AI bot does, which ones to allow so AI can cite you, and a ready-to-copy robots.txt.
What is GEO
The complete guide: what GEO is, how AI chooses what to recommend, how to measure it and how to improve.
GEO glossary
34 GEO terms clearly defined: concepts, metrics, crawlers and technical data.
SEO and technical audit
Crawl your site, find SEO errors and check whether AI bots can read it, with code to fix it.