What Is llms.txt?

llms.txt is a plain-text file you host at the root of your domain — just like robots.txt. But where robots.txt tells web crawlers how to index your pages, llms.txt tells large language models how to understand your brand.

The file communicates your company name, what you do in plain language, which pages contain your most authoritative content, your methodology or data, and anything else LLMs should know when citing you. When AI agents browse your domain or a model encounters your content, the llms.txt gives it structured context about who you are.

The spec was proposed by Jeremy Howard and the fast.ai team in 2024 as a lightweight, human-readable convention. It has no enforcement mechanism — LLMs are not "required" to read it — but Perplexity, several browse-enabled GPT configurations, and AI agents using RAG do actively use it to improve citation accuracy.

The llms.txt Spec Format

# Company Name
> A one or two sentence description of
> what your company does and who you
> serve.

## About
- [About page](https://example.com/about)
- [Team](https://example.com/team)

## Products
- [Product Name](https://example.com/product)
  Brief description of the product.

## Research & Data
- [Annual Report 2026](https://example.com/research)

## Methodology
- [How our score works](https://example.com/methodology)
  Explanation of data collection method.

## Optional
- [Blog](https://example.com/blog)

#H1: Your company name

>Blockquote: Your description (can be multi-line)

##H2: Section heading (About, Products, Research, etc.)

-List item: Page link + optional description

<\!-- Generator Tool -->

llms.txt Generator

Fill in the fields below. Your llms.txt generates live as you type.

<\!-- Input Form -->

Company Name *

Company Description *

Keep it factual and specific. Avoid marketing language.

Key Product / Service Pages

Use format: Title | URL — one per line

Blog / Research / Docs URLs

Include your highest-quality, most citable content

Key Methodology / Data to Cite

LLMs use methodology notes when citing your data in research contexts

Domain (optional)

Used to generate the canonical URL comment

<\!-- Output Panel -->

llms.txt

# Your Company Name
> Fill in the form to generate your llms.txt file.
> It will appear here as you type.

0 characters Download llms.txt

How to Host and Deploy Your llms.txt

Generate

Fill in the form above. Your llms.txt file generates live. Copy it to your clipboard or download the file directly.

Host at Root

Upload the file to your web server so it's accessible at https://yourdomain.com/llms.txt — the root of your domain, not in a subfolder.

Verify

Visit yourdomain.com/llms.txt in your browser. It should return the plain text file with Content-Type text/plain.

<\!-- Platform-specific instructions -->

Platform-Specific Setup

Vercel / Next.js

Place llms.txt in the public/ directory. Vercel serves all files in public/ at the root URL automatically. No configuration needed.

public/llms.txt → yourdomain.com/llms.txt

Express.js / Node

If you're using express.static, place the file in your static files directory. Or serve it explicitly:

app.get('/llms.txt', (req, res) => {'{'}
  res.setHeader('Content-Type', 'text/plain');
  res.sendFile(path.join(__dirname, 'llms.txt'));
{'}'});

WordPress

Upload llms.txt to the root of your web server (same directory as wp-config.php) via FTP or your hosting file manager. No plugin required.

Webflow / Framer

Webflow supports custom file publishing via the "Hosting" tab. Upload llms.txt as a static file. In Framer, use the custom domain's file publish feature.

Shopify

Add a page with the URL handle llms and place your content there. Note: Shopify appends .html by default — for a strict llms.txt you'll need a custom server-side redirect or use a Cloudflare Worker to serve the file.

<\!-- When LLMs pick it up -->

When Do LLMs Pick Up Your llms.txt?

Perplexity — Near Real-Time

Perplexity's RAG pipeline actively crawls and indexes llms.txt files. Once your file is live, Perplexity can pick it up within days during its next crawl of your domain. It uses the llms.txt to improve citation accuracy in its answer synthesis.

ChatGPT (Browse Mode) — Days to Weeks

When GPT-4o's browse mode is triggered and it navigates to your domain, it can read your llms.txt in that session. Consistent availability means your file is read whenever a browse query reaches your domain. Training-mode responses do not use llms.txt directly — that requires the next training cycle.

Claude — Session-Level (Agents)

Claude with web access reads llms.txt when it crawls your domain as part of an agentic task. The signal is used for citation context in that session. Claude's Opus and Sonnet models support the llms.txt spec for agent-initiated browsing tasks.

AI Agents and Crawlers — Immediate

Any AI agent using RAG that browses your domain will read llms.txt if they follow the convention. This includes custom GPTs, LangChain-based crawlers, and enterprise AI research tools. The convention is growing rapidly — early deployment means broad coverage before the field standardizes.

<\!-- agents.json relationship -->

llms.txt vs. agents.json: What's the Difference?

llms.txt and agents.json are complementary standards with different purposes. You should deploy both.

llms.txt

• Plain text — human and machine readable
• Describes your brand, content, and methodology
• Read by LLMs during browse sessions
• Improves citation accuracy and context
• Analogous to robots.txt — discovery convention

agents.json

• Structured JSON — machine-readable spec
• Defines what AI agents are permitted to do on your site
• Controls agentic access: read, transact, interact
• Used by AI agents for capability discovery
• Analogous to OpenAPI spec — capability declaration

Think of llms.txt as "here's who we are" and agents.json as "here's what you can do here." Together they form a complete LLM interface layer for your domain. See our agents.json →

<\!-- Best practices -->

llms.txt Best Practices

Keep descriptions factual and specific

LLMs use your description to form their understanding of your brand. Marketing superlatives ("the best," "the most powerful") are low-signal. Specific, factual statements ("measures citation frequency across 4 LLMs using 24 structured queries") are high-signal and more likely to be incorporated into accurate citations.

Link to your most citeable pages

Don't just list your homepage and product page. Include your research reports, statistics pages, comparison benchmarks, and methodology documentation — the content you actually want LLMs to cite. These are the pages that, when parsed, give LLMs the specific claims and data to include in their answers.

Describe your methodology explicitly

If you publish data, your methodology note is critical. LLMs cite data with methodology because human raters found methodologically transparent sources more credible. Explain how your numbers are collected, sample sizes, date ranges, and any limitations. A sentence of methodology dramatically increases citation probability for your statistics.

Keep it current

Update your llms.txt when you publish significant new research or products. An llms.txt pointing to a 2023 report when you have a 2026 edition published is wasting citation potential. Quarterly reviews are sufficient for most brands — monthly if you're publishing new research actively.

Serve with correct Content-Type

Your server must return Content-Type: text/plain for the llms.txt URL. Some CMS platforms default to text/html for any URL — this breaks the convention for parsers that check Content-Type. Verify with curl -I yourdomain.com/llms.txt.