Everyone Is Adding llms.txt for AI SEO. Google Says You Don't Need It.

That creates an interesting question:

If Google doesn't use it, why is everyone talking about it?

What is llms.txt?

llms.txt is a proposed standard for giving AI systems a curated map of important content on your website.

A simple example:

# Example.com
> Example is an AI SEO platform.

## Documentation
- [Getting Started](https://example.com/docs/start)
- [API Documentation](https://example.com/docs/api)
- [Pricing](https://example.com/pricing)

## Guides
- [AI SEO Guide](https://example.com/guides/ai-seo)
- [GEO Framework](https://example.com/guides/geo)

The idea is straightforward:

Make important information easier for AI systems to discover.

It sounds useful.

But there is an important distinction between a useful idea and a Google Search ranking signal.

Google is very clear about llms.txt

Google's current AI Search documentation explicitly lists llms.txt among things you don't need for visibility in Google Search or Google's generative AI features.

Google says Search doesn't use these special AI files.

It also says maintaining one for other services is fine, but it won't help or hurt your Google Search visibility or rankings.

That's a pretty strong statement.

So if someone tells you:

"llms.txt is a Google AI Overview ranking factor."

There is currently no basis for that claim.

But llms.txt Isn't Necessarily Useless

This is where things get more nuanced.

Google ignoring llms.txt doesn't mean every AI system ignores it.

The format was proposed specifically as a way for AI systems and agents to navigate important information on websites.

So there are two separate questions:

That's an important distinction.

Don't Confuse "AI-Readable" With "AI Ranking Signal"

This is one of the biggest problems with AI SEO advice.

Something can be designed for AI without being a ranking signal.

For example:

doesn't automatically mean:

Those are completely different claims.

You need evidence for the second one.

Think About What llms.txt Actually Does

Imagine a website has 5,000 pages.

An AI agent wants to understand the company.

Without a curated map, it might need to discover:

Homepage → Navigation → Categories → Articles → Documentation → Important pages

An llms.txt file could instead provide:

/llms.txt → Important pages → Relevant documentation → Key resources

That's potentially useful for discovery and navigation.

But discovery isn't the same as ranking.

And ranking isn't the same as citation.

There Are Actually Five Different Steps

When people say:

"AI uses my llms.txt."

Ask what they mean.

There are at least five different stages:

Published
   ↓
Crawled
   ↓
Parsed
   ↓
Used for retrieval
   ↓
Influences an answer
   ↓
Gets cited

A server log showing that an AI crawler requested /llms.txt only proves the first few steps.

It doesn't prove that the file influenced an answer.

And it certainly doesn't prove that it increased citations.

Google's AI Search Works Differently

Google explains that its generative AI features are grounded in its existing Search systems.

Google describes techniques including retrieval-augmented generation (RAG) and query fan-out.

In simplified terms:

User query
   ↓
Search / retrieval
   ↓
Relevant web sources
   ↓
Additional related searches
   ↓
Retrieved information
   ↓
AI-generated response

Google says these AI features rely on its core Search ranking and quality systems to retrieve relevant pages from its Search index.

That's why Google's advice is remarkably familiar:

Do good SEO.

Make your content:

• Crawlable
• Indexable
• Useful
• Original
• Reliable
• Well structured
• Easy for users to navigate

Google specifically says these fundamentals remain relevant for generative AI search.

What About Schema?

Google makes a similar clarification about structured data.

Structured data is useful.

Google uses it to understand page content and support certain Search features.

But Google says:

Structured data isn't required for generative AI search, and there is no special Schema.org markup you need to add.

So we now have an interesting pattern:

The takeaway isn't that structured data is useless.

It's that AI SEO isn't a collection of magic markup tricks.

Then Why Does llms.txt Matter?

It may still have a role in the broader agentic web.

AI agents are increasingly capable of navigating websites, reading pages, comparing information, and performing tasks.

Google itself is now discussing agent-friendly websites and emerging protocols for agentic experiences.

And other AI systems have their own crawling and discovery mechanisms.

For example, OpenAI documents OAI-SearchBot and recommends that publishers allow it to crawl content they want available to ChatGPT search experiences.

That makes a broader principle more useful:

Optimize your website for machines that actually consume your information — not for hypothetical AI signals.

The Better AI SEO Model

Instead of:

Think:

Technical SEO
   ↓
Quality content
   ↓
Clear entities
   ↓
Strong relationships
   ↓
Internal links
   ↓
External authority
   ↓
Crawlability
   ↓
Retrieval
   ↓
AI answer
   ↓
Citation

llms.txt is an optional layer that may help some systems discover your important resources.

It isn't the foundation.

Should You Create One?

There's no harm in experimenting with it.

If you have:

• Extensive documentation
• An API
• A large knowledge base
• A developer-focused product
• Lots of important resources
• AI agents interacting with your website

then a curated llms.txt could be a reasonable experiment.

But don't expect it to improve Google rankings.

And don't sell it to clients as a proven AI citation tactic.

The Real Lesson

The llms.txt story is actually useful because it teaches a bigger AI SEO lesson.

Don't confuse a proposed AI convention with a proven ranking signal.

Ask:

• Who uses it?
• Is that documented?
• What does the system actually do with it?
• Is there evidence that it changes visibility?

That's how you separate AI SEO engineering from GEO folklore.

The Bottom Line

llms.txt is an interesting idea.

It can act as a curated content map for compatible AI systems.

But for Google Search, the answer is clear:

So don't obsess over creating another file.

Build the thing that matters much more:

A website with useful content, clear entities, strong information architecture, accessible pages, consistent facts, and information that search and AI systems can actually retrieve.

llms.txt might become more important as the agentic web evolves.

But today, it's a convention to experiment with — not an AI SEO ranking hack.