What Is llms.txt and Does It Actually Work for SEO?

Someone told you to add an llms.txt file so ChatGPT, Perplexity or Google's AI answers would cite your site. Short answer: for most businesses, it doesn't move the needle today. Google says in writing that it ignores the file, and a study of 137,000 sites found that 97% of these files get read exactly zero times. It won't hurt to have one, but it's not the lever you're being sold. This guide explains what it is, why it's trending, and when it does make sense — in plain business language, not developer jargon.

What llms.txt is (in one sentence)

llms.txt is a text file placed at your site's root (yoursite.com/llms.txt) that lists your most important content as an index, so AI models can orient themselves without crawling your whole site. It was proposed in 2024 by Jeremy Howard, co-founder of Answer.AI, as a way for an AI that's already reading your site to quickly find what matters. That's all it was meant to be.

The original goal was technical and modest. The idea that it "boosts your AI visibility" came later — attached by the SEO industry on the speculation that AI platforms would reward the file. That's the part worth scrutinizing.

The question that matters: does any search engine or AI use it today?

This is the question your provider should answer before charging you to "optimize for AI". Here it is, with primary sources:

• Google: it ignores it, and says so in writing. In its official guide to appearing in AI features, in a section literally titled "Mythbusting", Google states you don't need files like llms.txt to appear in Search "as Google Search itself doesn't use them" and that doing so "will neither harm nor help your site's visibility or rankings in Google Search, as Google Search ignores them". Source: Google Search Central.
• John Mueller (Google) compared it to an obsolete tag. Google's advocate said no AI system currently uses llms.txt for search, calling it a "temporary crutch" useful only for AI coding tools parsing developer documentation. Source.
• OpenAI and the others control their access with robots.txt, not llms.txt. OpenAI's documentation explains how to manage its bots (GPTBot, OAI-SearchBot) through robots.txt; it doesn't mention llms.txt as a signal. Source: OpenAI.

And the hardest evidence: an Ahrefs study of 137,000 domains (May 2026) measured who actually reads these files. The result: 97% received no requests at all. Of the 3% that did get traffic, most came from SEO audit bots, not AI. The line that sums it up: "Slackbot alone fetched llms.txt files more often than PerplexityBot did". In other words, the AI bot that was supposed to benefit barely touches them. And crucially: no AI bot goes looking for an llms.txt that doesn't exist, so not having one leaves you off no radar.

A second study, by SE Ranking across roughly 300,000 domains, tested whether having the file translated into more citations in AI answers. It found no measurable relationship.

llms.txt is not the robots.txt of AI: what each one controls

Here's the most expensive misunderstanding. Many people believe llms.txt lets you control whether AI uses your content — to block or allow it. False: despite the name, it blocks nothing and allows nothing. It's an index, not a barrier. What actually controls AI access to your site is robots.txt, which has existed for decades.

In practice, if your goal is to stop being used to train models, that's done in robots.txt (for example, blocking GPTBot or enabling Google-Extended) — we cover it in detail in how to exclude your site from AI training and AI summaries. And if you want to appear in ChatGPT's search, you have to allow its search bot (OAI-SearchBot). llms.txt plays no part in either decision.

What actually decides whether AI reaches your site

While llms.txt dominates the conversation, the standard that's actually consolidating to control AI lives in robots.txt. On October 30, 2026, Google announced support for a "shared AI training control": a token inside robots.txt, based on an IETF standard (the body that sets internet rules), that lets you express in one place whether you want AI providers to use your content for training. Google plans to support it, alongside other providers, in 2027.

Translated to business terms: the industry is standardizing AI control on robots.txt, not on llms.txt. If someone wants to sell you "preparing your site for AI", this is the real conversation — not the trendy file.

So, should you add one? It depends on your business

It's not black and white. It depends on what kind of site you run:

• Online store, local business, corporate or services site → not a priority. For you, llms.txt is decoration today: it won't earn you ChatGPT citations or Google rankings. If your CMS generates it for free, leave it; but don't pay to implement it or treat it as a growth lever. Your time pays off more elsewhere (see the next section).
• Software company, SaaS, or a product with technical documentation → here it does make sense. The only audience the data shows reading these files are coding agents (the kind developers use with Claude or ChatGPT to write code). If your customers integrate your product or read your docs with those tools, a well-built llms.txt of your documentation can help. Not by coincidence, OpenAI and Anthropic publish an llms.txt… for their own developer documentation, which is exactly this case.

One detail almost nobody mentions: because AI agents are designed to trust what the llms.txt says, if your site auto-generates it you should review it. Researchers are already studying it as a possible manipulation vector. A file you don't control is a file you shouldn't publish blindly.

What to do today to get found by AI

If your real goal is for AI to talk about your business — cite you, recommend you, show you in a summary — the good news is there's no file trick. What works is, largely, good SEO. Google's own guide for generative AI says it plainly: its AI search relies on the same ranking systems as always. In order of impact:

• Original content with a point of view, not recycled. AI answers look for sources with a distinct perspective. A page that repeats what's already everywhere doesn't get cited; one that adds real experience or data does.
• Make your site technically readable: indexable, fast, with a clear structure and clean HTML. If Google can't understand your site, its AI can't either.
• Direct answers and clear structure: headers phrased as questions, answers in the first lines, tables and concrete definitions. That's how AI extracts and cites — the basis of answer engine optimization.
• For local businesses and ecommerce: your Google Business Profile and, if you sell, your Merchant Center feed weigh more than any text file.
• Access control lives in robots.txt: consciously decide which AI bots you allow and which you don't, depending on whether you want to appear on each platform.

Bottom line: before chasing the trendy file, cover what actually decides your visibility. If you'd like us to review how AI sees your business — or doesn't — today, that's exactly what we work on at SEO for AI.

Frequently asked questions

If I add an llms.txt, will ChatGPT or Perplexity cite me?
Not because of that. The bots that generate citations in AI answers barely read these files, and studies found no relationship between having llms.txt and getting more citations. What makes you citable is useful, original, well-structured content — we break it down in who ChatGPT cites and why.

Is llms.txt like robots.txt but for AI?
No. robots.txt controls which bots enter your site; llms.txt blocks and allows nothing — it's just an index. If you want to control AI access, that's done in robots.txt.

Does Google use llms.txt for its AI summaries (AI Overviews)?
No. Google states in writing that it ignores the file and that it affects your rankings neither for better nor worse.

Should I have one anyway, just in case?
If your platform generates it at no cost, it doesn't hurt to have it. But don't pay to implement it or expect results: for most businesses it's neutral today. The exception is products with technical documentation for developers.

How do I control whether AI uses my content?
With the robots.txt file: you can allow or block specific bots like GPTBot (OpenAI training) or Google-Extended (Google's AI). In 2027, a shared control — also in robots.txt — is being added to express that preference across several providers at once.