llms.txt Explained: Does Your SaaS Website Need One?

You saw it on X. Someone shipped an llms.txt file, posted a screenshot, and called it "the new robots.txt for AI." A dozen replies later, half the thread swears it's essential for 2026 and the other half calls it snake oil. So which is it?

Here's the short version: llms.txt is real, it's easy to add, and most of the noise about it is wrong in both directions. It won't 10x your AI visibility. It also won't hurt you. The honest answer to "does your SaaS website need one" is it depends on what you're trying to do, and by the end of this post you'll know exactly which camp you're in.

Let's cut through it.

What is llms.txt?

So, what is llms.txt? It's a plain-text file you place at the root of your domain, at https://yoursite.com/llms.txt (for example, avinashvagh.com/llms.txt), written in Markdown. Its job is to give large language models a clean, curated map of your most important content, with a one-line description next to each link.

Think of it as a tour guide for AI. Instead of an AI crawler parsing your full HTML page, navigation, cookie banners, JavaScript, ads, the works, the llms.txt file points it straight to the pages that matter and tells it what each one is for.

The format was proposed in September 2024 by Jeremy Howard of Answer.AI, the same person behind fast.ai. The full spec lives at llmstxt.org. A basic file needs just one required element: a single H1 title with your project or site name. Everything after that, a short summary, grouped link sections, an optional low-priority section, is there to add structure.

There's also a sibling file you'll hear about: llms-full.txt. The difference between llms.txt and llms-full.txt is scope. The standard llms.txt is a slim outline of links and descriptions. The llms-full.txt version inlines the actual content of those pages into one long document, so an AI can ingest everything in a single fetch. Small docs sites sometimes ship both.

That's the whole concept. A curated, machine-readable index, sitting at your root directory, written for AI instead of humans.

llms.txt vs robots.txt: not the same job

This is the comparison everyone reaches for, and it's where most of the confusion starts. The llms.txt vs robots.txt question has a clean answer once you separate access from understanding.

robots.txt controls access. It's a decades-old standard that tells crawlers which URLs they're allowed to fetch and which to skip. It uses User-agent, Allow, and Disallow directives. It's a gatekeeper, useful for keeping bots out of /admin/ or your staging folder.

llms.txt controls understanding. It doesn't block anything. It does the opposite: it surfaces your best content and explains it, so AI models don't have to guess what matters on your site.

A quick way to hold the difference in your head:

• robots.txt points to "Here's what you may and may not crawl."
• sitemap.xml points to "Here's every page so you can discover them all."
• llms.txt points to "Here are the pages that actually matter, and here's what each one is about."

They don't compete. A well-run SaaS site can have all three. One rule worth knowing: never Disallow your /llms.txt or your Markdown files inside robots.txt, or you'll block the very file you just built.

Does llms.txt actually work? The honest part

Here's where most articles get cheerful and vague. We won't. If you're shipping llms.txt to boost AI search visibility, more citations in ChatGPT, Claude, Perplexity, or Google's AI Overviews, the evidence right now is underwhelming, and you deserve to hear it straight.

Adoption is low. A SE Ranking study of 300,000 domains found an llms.txt adoption rate of about 10%. After more than 18 months of industry chatter, roughly one site in ten has the file. Among the most AI-cited domains in that dataset, almost none had one.

The big AI crawlers mostly ignore it. Independent crawler-log studies tell the same story. One analysis tracked tens of thousands of AI bot visits over 90 days and found that a tiny fraction, well under 1%, actually requested the llms.txt file. GPTBot, ClaudeBot, PerplexityBot, and Google-Extended overwhelmingly skip it and crawl your HTML directly.

Google has said no, on the record. In mid-2025, Google's Gary Illyes stated Google does not use llms.txt and isn't planning to. John Mueller went further and compared it to the old meta keywords tag, a signal people optimized for that search engines quietly ignored. As of early 2026, no major provider, OpenAI, Anthropic, Google, Meta, Mistral, has publicly committed to using llms.txt as a ranking or citation signal in production.

So does llms.txt work? Not as the SEO lever it's being sold as. If someone promises you ranking gains from a text file, raise an eyebrow.

But, and this matters, that's not the only reason to ship one.

Where llms.txt genuinely matters

The strongest use case for llms.txt has never been broad search visibility. It's usability for AI agents that already know about you.

When a developer points an AI coding assistant at your docs, or an agent fetches your API reference mid-task, a clean llms.txt file gives it exactly the right pages instead of a wall of HTML noise that burns through the context window. This is a Business-to-Agent (B2A) play, not an SEO play. You're publishing a machine-readable surface that AI tools can route on once they're already pointed at you.

That's why the companies getting real value tend to be developer-facing. Who uses llms.txt today? Anthropic, Stripe, Cursor, Vercel, Mintlify, and Supabase, among others, teams whose users routinely feed their docs into AI assistants. Vercel even adds contextual descriptions so agents pick the right API endpoints.

If your SaaS has documentation that developers paste into Claude or Cursor, llms.txt earns its keep. If you're a five-page marketing site hoping for more ChatGPT mentions, it mostly won't move the needle, yet.

llms.txt example you can copy

Talk is cheap, so here's a working llms.txt example. This is roughly what a developer-tools SaaS would ship:

# Acme API

> Acme is a payments API for SaaS apps. This file points AI tools to our
> most useful documentation and reference pages.

## Docs

- Quickstart: Get your first payment flowing in 10 minutes.
- Authentication: API keys, OAuth, and webhook signing.
- API Reference: Every endpoint with request and response examples.

## Guides

- Webhooks: Set up and verify event notifications.
- Testing: Sandbox keys and test card numbers.

## Optional

- Changelog: Recent API changes.
- Pricing: Plans and limits.

Use that as your llms.txt template and swap in real links. A few rules that separate a useful file from a junk one:

• Keep it short. Curate 20 to 50 links, not your whole sitemap. This is a greatest-hits list, not a duplicate.
• Write descriptions for context. "This explains our pricing tiers" beats keyword soup like "affordable enterprise SaaS pricing solution."
• Link to Markdown versions of pages where you have them, they're cleaner for AI to ingest than HTML.
• Use the ## Optional section for lower-priority pages so models can skip them when context is tight.
• Skip gated content. AI can't read pages behind a login anyway.

How to create llms.txt (and where to put it)

You've got two paths, depending on how much you want to touch.

By hand. Open a text editor, follow the format above, save it as llms.txt, and upload it to your site's root directory so it resolves at yoursite.com/llms.txt. It has to live at the root, subfolders won't register. That's the whole job for most sites.

With a generator or plugin. If writing Markdown by hand isn't your thing, our free llms.txt generator scaffolds one from your sitemap, fetches your page titles and descriptions, and lets you trim from there. On WordPress, an llms.txt WordPress plugin can auto-generate the file and even track whether GPTBot or ClaudeBot are accessing it. Docs platforms increasingly handle it natively, GitBook and Mintlify generate llms.txt for published docs automatically, so there's nothing to maintain.

One caution on the "set it and forget it" plugins: a stale llms.txt that points to deleted or outdated pages works against you. Build a quick quarterly review into your content calendar, and update the file whenever you launch major pages or restructure your docs.

Want to know whether AI crawlers are even reaching your site before you bother? You can check your site with our free AI Crawler Checker and see which bots are actually showing up in the first place.

Do you need llms.txt for your SaaS? A simple decision

Strip away the hype and the "is llms.txt worth it" question comes down to a 30-second gut check.

Ship one if:

• You have developer docs or an API reference people feed into AI tools
• Your team can keep it updated without it becoming a chore
• You think of it as future-proofing and agent usability, not a ranking hack

Skip it (for now) if:

• You're a small marketing site expecting AI traffic to jump
• You'd have to fake a reason to create one
• You don't have docs worth curating yet

For most developer-facing SaaS teams, the math is easy: the file takes an hour, the downside is zero, and the upside is real for the AI tools your users already run. Treat it as cheap insurance, not a growth channel.

And if your real goal is getting cited in AI answers, llms.txt is the wrong tool, that's a content and authority game. Our complete guide to answer engine optimization covers what actually moves AI visibility.

Bottom line on llms.txt

llms.txt is a low-effort, low-risk file that gives AI tools a clean map of your best content. It is not an SEO ranking factor, Google doesn't use it, and the major AI crawlers rarely fetch it today. But for developer-facing SaaS with real documentation, it's a sensible piece of the AI-readiness stack, and being early costs almost nothing if the standard does gain ground.

Don't ship it because a viral thread told you to. Ship it because your docs deserve a clean front door for the agents already knocking.

Ready to build yours? Generate one in seconds with our free llms.txt generator, or check who's actually crawling your site first with the AI Crawler Checker.