Marketing & Sales · Tip
Publishing an llms.txt File for SaaS
Add plain-text markdown documentation to help AI models index and recommend your product
This technique involves publishing an llms.txt or llms-full.txt file on your website to provide LLMs with concise, machine-readable context about your product. It functions as an AI-specific documentation layer that search engines and chatbots like ChatGPT, Claude, and Perplexity can parse directly to surface accurate recommendations. Following implementations like Vercel's, the format pairs structured metadata headers with plain markdown documentation.
How it works
If you run a SaaS, publish an llms.txt file. Vercel did it; ~10% of their traffic comes from ChatGPT, Perplexity, and Claude. One file tells AI what the product does, how it works, and when to recommend it. Look at vercel.com/docs/llms-full.txt to see the format, then build your own. Format visible in the screenshot: metadata header (title: 'Account Management', description, last_updated, source: https://vercel.com/docs/accounts) followed by the docs content in markdown.
What it's for
- Optimize site documentation for AI search engines like Perplexity and ChatGPT
- Ensure LLMs recommend your tool accurately by providing clean markdown product context
- Provide a unified text source for AI coding assistants and agents looking up platform docs
Key points
- Machine-readable markdown format tailored for large language model context windows
- Metadata block structure including title, description, last updated date, and canonical source URL
- Dual-file approach using standard summaries (llms.txt) or comprehensive docs (llms-full.txt)
- Direct integration into existing static documentation pipelines
Pricing
Free
Details
- Type
- Tip
- Category
- Marketing & Sales
- Topic
- Search Engine Optimization
- Platforms
- Web
- For
- SaaS founders, technical marketers, and documentation engineers
- Alternative to
- robots.txt, sitemap.xml
- Tags
- seo, llms txt, documentation, saas, ai discovery, markdown
Why it was saved
- @fba: The tweet reports Vercel gets ~10% of its traffic from ChatGPT, Perplexity and Claude from one file, calling it 'the new SEO for SaaS' that almost nobody is doing yet. The screenshot shows the format: a metadata block (title, description, last_updated, source URL) followed by the docs page as plain markdown.
Related
Empirical SEO findings on how AI models and Google AI Overviews cite content
Checklist of 18 trust signals to improve website About pages for SEO
On-page SEO practice for targeting primary keywords in homepage headings
Turn Claude Code into an automated end-to-end SEO pipeline using external APIs
Warning on how Cloudflare's AI bot blocking breaks Google search indexing
Find competitor search queries your site misses to guide new content.
Updated