What Is llms.txt and Why Every B2B SaaS Site Needs One

In 2024, a new informal standard emerged from the AI developer community: llms.txt.

The idea is simple. AI engines crawl the web the way search engines do, but they consume content differently — they're looking for text they can synthesise, summarise, and cite in responses, not just index for keyword ranking. A file at yoursite.com/llms.txt gives these models a curated, structured starting point: here's what this product is, here's who it's for, here's the content most worth reading.

It's one of the cheapest, fastest ways to improve how AI engines understand and describe your brand. And the majority of B2B SaaS companies don't have one.

The Problem llms.txt Solves

When an AI engine crawls your site without guidance, it faces several challenges:

Content priority. Your domain has hundreds or thousands of pages — blog posts, help docs, landing pages, legal pages. Without a signal about what matters most, the crawler may spend its crawl budget on changelog entries and cookie policy pages rather than your core product content and case studies.

Entity disambiguation. If your company name is a common word or acronym, the crawler may conflate your brand with other entities. llms.txt provides explicit context: "We are [Company], a [category] platform for [buyer type]."

Structure over marketing copy. Your homepage is written for humans making emotional and rational decisions. llms.txt is written for models that want factual, structured, citation-ready descriptions. The tone and format can be radically different — and more useful.

The llms.txt Format

The file lives at /llms.txt on your root domain (alongside robots.txt and sitemap.xml). It uses a structured Markdown format with specific sections:

# [Your Product Name]

> [One sentence: what it is, who it's for, what makes it different.]

[2-3 paragraphs of factual product description. Write for an AI that needs
to accurately describe your product in a buyer query response. Be specific
about category, use cases, buyer segments, and key differentiators. Avoid
marketing superlatives.]

## Key Pages

- [Homepage](https://yoursite.com): Product overview
- [Pricing](https://yoursite.com/pricing): Pricing tiers and plan details
- [Integrations](https://yoursite.com/integrations): Integration marketplace
- [Case Studies](https://yoursite.com/customers): Customer outcomes with metrics
- [vs Competitor A](https://yoursite.com/vs-competitor-a): Comparison page
- [vs Competitor B](https://yoursite.com/vs-competitor-b): Comparison page

## Product Details

- **Category:** [G2 category name]
- **Pricing:** Starting at $[X]/month per [unit]
- **Deployment:** SaaS / Cloud-hosted
- **Primary buyers:** [Job titles and company sizes]
- **Key integrations:** [List top 5-10 integrations]
- **G2 rating:** [X.X]/5 from [N] reviews — [Badge if applicable]

## Documentation

- [API Reference](https://docs.yoursite.com/api)
- [Getting Started](https://docs.yoursite.com/quickstart)
- [Changelog](https://yoursite.com/changelog)

What to Write in Each Section

The opening summary is the most important part. This is what AI engines will use to describe your product when they don't have access to your full site. Write it as if you're answering: "What is [Product] in two sentences?" Be specific about category (not "a platform" — say "a customer success platform"), buyer type (not "businesses" — say "mid-market B2B SaaS companies with 50–500 customers"), and key differentiator.

Key pages should prioritise content that AI engines can cite in buying queries: pricing, comparison pages, case studies with metrics, and integration listings. Don't just link your blog homepage — link to specific high-value posts.

Product details provide the structured facts that make AI responses accurate. The G2 rating field is particularly important: it gives the model a citable social proof claim it can use with confidence.

Documentation links are valued by Claude and Perplexity in particular for technical evaluation queries — buyers asking about API capabilities, integration options, or implementation requirements.

Claude's MCP Marketplace and llms.txt

For B2B SaaS companies building AI-adjacent products, there's an additional layer beyond llms.txt: Claude's MCP (Model Context Protocol) marketplace.

Claude surfaces brands that appear in its MCP marketplace in responses to queries about AI-native tools and integrations. If your product has or could have an MCP server — a structured interface that allows Claude to interact with your product — listing it in the marketplace gives you a citation surface in Claude-specific buying queries that most competitors haven't thought about yet.

This isn't relevant to every B2B SaaS product. But for companies in categories adjacent to AI workflows — project management, CRM, data platforms, analytics, documentation — it's an emerging high-value signal that will compound as Claude's user base grows.

Implementation Checklist

1. Create /llms.txt at your domain root. Make sure it's server-side rendered and not blocked by robots.txt.
2. Write the opening summary — 2-3 paragraphs, factual, category-specific, buyer-type-specific.
3. Link your five highest-value pages — pricing, comparison pages, case studies with metrics.
4. Add product facts — G2 rating, pricing floor, key integrations, buyer personas.
5. Submit to Perplexity's source index — Perplexity allows direct source submission via their publisher portal.
6. Check robots.txt to ensure GPTBot, ClaudeBot, PerplexityBot, and OAI-SearchBot can reach /llms.txt.

The file takes two to four hours to write well. The citation benefit — more accurate AI descriptions of your product across all major engines — compounds over every crawl cycle indefinitely.

---

This post is adapted from Chapter 11 of The Citation Economy — the playbook for B2B SaaS AI visibility.