You've optimized for Google. You've set up your robots.txt. Maybe you've even added structured data. But there's a new file format that's quietly becoming important for AI visibility, and most websites don't have it yet.
It's called llms.txt. Think of it as a cheat sheet for AI models. A plain-text file that sits at the root of your site and tells language models what you do, which pages matter, and how your content is organized.
Here's how to create one, step by step.
What is llms.txt and why should you care?
An llms.txt file is a standardized way to communicate with AI models about your website. It lives at yoursite.com/llms.txt, just like robots.txt lives at your domain root.
The idea is straightforward. When an AI crawler or language model visits your site, it can read this file to quickly understand your site's purpose, structure, and most important content. Instead of making the AI figure it out from thousands of pages, you hand it a curated summary.
The specification comes from llmstxt.org and uses a simple markdown format. It's human-readable, easy to create, and lightweight.
Why does this matter now? Because AI search is growing fast. ChatGPT, Perplexity, Claude, Gemini, and Grok are all answering questions that used to go to Google. If these models can understand your site quickly and accurately, you have a better shot at being mentioned when someone asks a question relevant to your business.
The llmstxt.org specification explained
The format is intentionally simple. Here's the structure, section by section.
1. An H1 title (required). This is the only mandatory element. It's the name of your site or project, written as a markdown heading.
2. A blockquote summary (optional but recommended). A short paragraph introduced with > that captures the essential information about your site. Think of it as your elevator pitch for AI.
3. Additional context paragraphs (optional). Any extra detail you want AI to know. This can include what you do, who you serve, what makes you different. No headings here, just plain text or lists.
4. Sections with file lists (optional). These are H2 headings followed by markdown lists of links. Each link follows this syntax: [Page Name](URL): Brief description of what this page covers. This is where you point AI to your most important pages.
5. An "Optional" section (reserved name). If you include a section titled "Optional", it signals that those URLs can be skipped when the AI needs a shorter context. Use it for secondary or supplementary content.
That's it. No complex syntax, no special encoding, no configuration files. Just clean markdown.
Creating an llms.txt file manually (step by step)
Let's walk through building one from scratch for a fictional SaaS company called "InvoiceFlow" that makes invoicing software for freelancers.
Step 1: Write your H1 title
Start with your site or brand name:
# InvoiceFlow
Step 2: Add a blockquote summary
Describe what you do in 1-3 sentences. Be specific. Include your target audience and your core value proposition.
> InvoiceFlow is invoicing software built for freelancers and solo consultants. It automates recurring invoices, tracks payments, and handles tax calculations for independent professionals in the US and Europe.
Step 3: Add context paragraphs
This is where you can expand with detail that helps AI understand your positioning. Keep it factual and concise.
InvoiceFlow serves over 12,000 freelancers across 30 countries. The platform integrates with Stripe, PayPal, and major European payment providers. Pricing starts at $9/month.
Step 4: Create sections with your key pages
Think about the pages you most want AI to know about. Group them logically under H2 headings.
## Product
- [Features](https://invoiceflow.com/features): Complete feature overview including automated invoicing, payment tracking, tax calculations, and client management.
- [Pricing](https://invoiceflow.com/pricing): Three plans from $9/month (Solo) to $49/month (Studio). All plans include unlimited invoices.
- [Integrations](https://invoiceflow.com/integrations): Connects with Stripe, PayPal, QuickBooks, Xero, and 20+ other tools.
## Resources
- [Blog](https://invoiceflow.com/blog): Guides on freelance invoicing, tax tips, and getting paid faster.
- [Help Center](https://invoiceflow.com/help): Product documentation and tutorials.
- [API Documentation](https://invoiceflow.com/docs/api): REST API reference for developers building on InvoiceFlow.
## Company
- [About](https://invoiceflow.com/about): Founded in 2023 by former freelancers. Bootstrapped and profitable.
- [Contact](https://invoiceflow.com/contact): Support and partnership inquiries.
## Optional
- [Changelog](https://invoiceflow.com/changelog): Product updates and release notes.
- [Terms of Service](https://invoiceflow.com/terms): Legal terms.
- [Privacy Policy](https://invoiceflow.com/privacy): Data handling and privacy practices.
Step 5: Put it all together
Your complete llms.txt file looks like this:
# InvoiceFlow
> InvoiceFlow is invoicing software built for freelancers and solo consultants. It automates recurring invoices, tracks payments, and handles tax calculations for independent professionals in the US and Europe.
InvoiceFlow serves over 12,000 freelancers across 30 countries. The platform integrates with Stripe, PayPal, and major European payment providers. Pricing starts at $9/month.
## Product
- [Features](https://invoiceflow.com/features): Complete feature overview including automated invoicing, payment tracking, tax calculations, and client management.
- [Pricing](https://invoiceflow.com/pricing): Three plans from $9/month (Solo) to $49/month (Studio). All plans include unlimited invoices.
- [Integrations](https://invoiceflow.com/integrations): Connects with Stripe, PayPal, QuickBooks, Xero, and 20+ other tools.
## Resources
- [Blog](https://invoiceflow.com/blog): Guides on freelance invoicing, tax tips, and getting paid faster.
- [Help Center](https://invoiceflow.com/help): Product documentation and tutorials.
- [API Documentation](https://invoiceflow.com/docs/api): REST API reference for developers building on InvoiceFlow.
## Company
- [About](https://invoiceflow.com/about): Founded in 2023 by former freelancers. Bootstrapped and profitable.
- [Contact](https://invoiceflow.com/contact): Support and partnership inquiries.
## Optional
- [Changelog](https://invoiceflow.com/changelog): Product updates and release notes.
- [Terms of Service](https://invoiceflow.com/terms): Legal terms.
- [Privacy Policy](https://invoiceflow.com/privacy): Data handling and privacy practices.
That took about 10 minutes. And it gives any AI model a clear, structured understanding of what InvoiceFlow is, what it does, and where to find the important content.
The easy way: use a free generator
Writing llms.txt manually works fine for small sites. But if you have dozens or hundreds of pages, you don't want to type out every link by hand.
Mentionable offers a free llms.txt generator that does the work for you. Enter your URL, and it scans your sitemap, discovers your pages, organizes them into logical sections, and produces a spec-compliant file you can download and deploy. No signup required.
It's particularly useful if you're not sure which pages to include or how to organize them. The generator analyzes your site structure and makes sensible grouping decisions automatically.
Where to place your file and how to verify it works
Deploying llms.txt
Place the file at the root of your domain so it's accessible at:
https://yoursite.com/llms.txt
How you deploy depends on your setup:
Static sites (Next.js, Gatsby, Hugo, etc.): Drop the file in your public/ folder. It'll be served automatically at the root.
WordPress: Upload llms.txt to your root WordPress directory (same level as wp-config.php), or use an .htaccess rule to serve it.
Managed platforms (Webflow, Squarespace, Shopify): Check if your platform supports custom files at the root. Some do natively, others require workarounds via custom code or redirects.
Verifying it works
After deployment, open https://yoursite.com/llms.txt in your browser. You should see the raw markdown content. If you get a 404, the file isn't in the right place.
Check that:
- The file is served with a
text/plainortext/markdowncontent type - No authentication is required to access it
- The formatting looks correct (headings, links, blockquote all rendering as expected in raw form)
Best practices and common mistakes
Do this
Be specific in your descriptions. "SaaS invoicing tool for freelancers" is better than "business software." The more precise you are, the more accurately AI can represent you.
Prioritize your pages. Don't list every page on your site. Focus on the 10-30 pages that best represent what you do and what you want to be known for.
Use the "Optional" section wisely. Legal pages, changelogs, and supplementary content belong here. AI can skip them when context space is limited.
Keep it current. When you add a major new page, update your llms.txt. When you remove a page, take it out. A quarterly review is a good habit.
Write descriptions that stand alone. Each link description should make sense without clicking the link. "Complete pricing comparison of all three plans including annual discounts" is better than "Pricing page."
Don't do this
Don't list hundreds of pages. The point is curation, not completeness. Your sitemap already lists every page. llms.txt highlights the ones that matter.
Don't use marketing fluff in descriptions. "The world's most innovative solution" tells AI nothing. "Invoicing software for freelancers, starting at $9/month" tells it everything.
Don't forget to update it. A stale llms.txt with broken links or outdated descriptions is worse than having none at all.
Don't include pages you don't want AI to reference. If you have internal pages or outdated content, leave them out. Only include what you'd want AI to cite.
What to expect (honest talk about adoption)
Let's be real about where things stand.
The llms.txt standard is still emerging. There's no guarantee that ChatGPT, Claude, or Perplexity actively reads your llms.txt file today. The spec exists, adoption is growing, but it's not universally implemented by AI providers yet.
So why bother?
First, the cost is near zero. You can set up an llms.txt file in 10 minutes. Even if no AI reads it right now, you've lost almost nothing.
Second, adoption is trending upward. More sites are adding llms.txt files. AI companies are aware of the specification. When broad adoption reaches a tipping point, you'll already be ready.
Third, the exercise itself is valuable. The act of curating your most important pages and writing clear descriptions forces you to think about how AI sees your site. That clarity helps your broader AI visibility strategy, regardless of whether llms.txt is read by models today.
Think of it like schema markup five years ago. The sites that added it early didn't see immediate results. But when Google started using it heavily for rich snippets, those sites had a head start.
Beyond llms.txt: your full AI visibility picture
An llms.txt file is one piece of the puzzle. It helps AI understand your site structure, but it doesn't tell you whether AI models actually mention you in their responses.
For that, you need to track your visibility across the AI platforms your customers use. Are you showing up when someone asks ChatGPT for a recommendation in your category? Do Perplexity and Claude cite you as a source?
If you're curious about where you stand right now, run a free GEO audit to check your site's AI readiness, including bot access, structured data, and content structure. It takes 30 seconds and gives you a concrete starting point.
Creating your llms.txt file is a solid first step. Knowing whether AI actually recommends you is the next one.
