You've probably seen it lately: «llms.txt». It's showing up in articles, LinkedIn posts, marketing newsletters. And if nobody's mentioned it to you yet, give it a few days. Someone will come along and tell you that you need it. That it's urgent. That your site is incomplete without it. And the truth is they're partly right — but only partly.
Let me get to the point.
What exactly is this llms.txt thing?
An llms.txt file is, at its core, a plain text file that lives at the root of your website — just like the robots.txt you're probably already familiar with — and that describes, in a curated summary, what your site is about and which pages matter most. Its stated purpose is to give an AI a direct map of your content, instead of forcing it to crawl page by page trying to figure out what you do and who you do it for.
Think of it this way: an AI arriving at your site is like a new reader with no context, no history, no idea whether you're a B2B consultancy or a clothing store. The llms.txt is like handing that reader a business card before they start reading. «Hi. This is what I do. This is who I do it for. Here are my most relevant pages.» Clean, direct, no fluff.
It's not an official standard set by any governing body. It's a community-driven proposal that several auditing tools already factor in when scoring how «AI-readable» your site is. And that's exactly what makes it interesting — and also what puts it squarely in a pretty large gray area.
The opinion nobody asked for:
The major AI systems — ChatGPT, Claude, Perplexity, Google AI Overviews — have not publicly confirmed that an llms.txt file is a determining factor in whether they cite your site or not. As of this writing, there's no clear statement from any of them on that point. And in a space where the rules change every three weeks, that matters more than it might seem.
So what does it actually do? About what many things in SEO and GEO — Generative Engine Optimization — do right now: there's no certainty it works, but no certainty it doesn't either. And when something is free, takes thirty seconds to implement, and causes no harm, the conversation simplifies considerably.
The real problem isn't the llms.txt. The real problem is that most companies are out there looking for that kind of magic lever — the file, the plugin, the trick — when the underlying work that actually determines whether an AI cites you is considerably deeper and considerably less glamorous.
SEO, GEO, and AEO: three distinct disciplines that most people mix together
It's worth pausing here for a moment, because the llms.txt lives at the intersection of three disciplines that get confused constantly.
Traditional SEO — Search Engine Optimization — is the practice of optimizing your site to appear in Google's search results. Rankings, clicks, organic traffic. What you've been hearing about for years.
GEO — Generative Engine Optimization — is something different: it's the practice of optimizing your content so that generative AI models, like ChatGPT or Perplexity, recognize you as an authoritative source and cite you in their responses. You're not chasing a spot in a ranking. You're trying to get the AI to mention you when someone asks something related to your industry.
AEO — Answer Engine Optimization — takes things a step further in a different direction: optimizing your content to be the direct answer to a question, without the user needing to go anywhere else. Google's featured snippets, AI Overviews, voice assistants. The goal is to be the answer — not just to appear near it.
You can check our SEO/GEO/AEO glossary for more detailed definitions, but the practical distinction is this: SEO puts you on Google's map, GEO puts you in the AI's mouth, and AEO turns you into the answer nobody needs to search for twice.
| SEO | GEO | AEO | |
|---|---|---|---|
| Goal | Rankings on Google | Be cited by generative AI | Be the direct answer |
| Main engine | Google Search | ChatGPT, Perplexity, Claude | Google AI Overviews, voice assistants |
| Key metric | Rankings, organic traffic | Mentions, citations in AI responses | Appearances as direct answer |
| Role of llms.txt | Indirect / minimal | Possibly useful, unconfirmed | Possibly useful, unconfirmed |
The llms.txt theoretically lives in GEO and AEO territory. But — and here's the uncomfortable part — having a flawless llms.txt on a site with no real authority, no content that answers specific questions, no properly implemented structured data, and no AI crawler access... is like putting a nice sign on the door of a house that hasn't been built yet.
What you can actually do today, and why it's worth doing
I just added an llms.txt generator to AEON42, the SEO/GEO/AEO platform we built here at Play.Interactive. I did it for a simple reason: if something is free, takes thirty seconds, and causes no harm, there's no logical argument for not having it. The generator takes your sitemap, processes it, and gives you a ready-to-review file. You don't need to know how to code. You don't need to understand the technical spec in depth. A couple of minutes and it's done.
But I also added four other tools at aeon42.com/herramientas — all free, all without registration, all the kind that actually matter:
- A full site audit: speed, structure, technical errors.
- A check of whether ChatGPT, Claude, and Perplexity crawlers can actually access your pages — because many sites block them without knowing it.
- A structured data inspector (schema JSON-LD), which is what actually helps AIs understand what your content is about.
- A tool that detects whether Google is already citing you in its AI Overview responses — or who's appearing in your place.
No credit card. No registration. These are tools, not sales funnels dressed up as tools.
The order matters more than the checklist
Generating your llms.txt without first checking whether AI crawlers can read your site is like writing a great cover letter without knowing if the envelope will arrive. Check crawler access first, then review your structured data, then generate the llms.txt. In that order, things actually make sense.
Why a blocked crawler is the problem nobody's looking at
Of all the tools I mentioned, the crawler check is the one that produces the most surprises. I keep seeing sites with reasonably solid technical SEO — decent speed, clean URL structure, even schema markup — that have ChatGPT's GPTBot, ClaudeBot, or PerplexityBot blocked in their robots.txt, and nobody at the company knows it. Sometimes a previous agency configured it two years ago. Sometimes a security plugin turned it on automatically. Sometimes nobody ever checked.
If AI crawlers can't read your site, the llms.txt is useless. The structured data is useless. All the content work you've done reaches nowhere from an AI visibility standpoint. It's the first item on the list, and also the most frequently overlooked.
If you want to understand how this entire AI visibility process works and what a complete strategy looks like, our SEO, GEO, and AEO service page breaks it down in detail.
The gray area, stated honestly
There's something I think is worth saying plainly, because there's a lot of motivated optimism circulating on this topic: nobody knows with certainty how these factors are weighted in AI models' decisions. Nobody. The rules change. Models get updated. What works today may stop working in six weeks, and what seems irrelevant today may become critical tomorrow.
What we do know is that there's a set of things that clearly helps: content that answers specific questions directly, well-implemented structured data, open access for the right crawlers, and a consistent brand presence across multiple sources. The llms.txt is, at best, an additional layer on top of that work. It doesn't replace it.
What strikes me as dishonest — and I'm seeing it from several services and agencies that have appeared in the last few months — is selling someone an llms.txt as the solution to their AI visibility problem. It's like selling the sign without the store.
What actually makes sense to do this week
Verify that AI crawlers can read your site
Before anything else. If ChatGPT's GPTBot, ClaudeBot, or PerplexityBot are blocked in your robots.txt, no downstream optimization will land anywhere. The free tool at aeon42.com/herramientas checks this in seconds.
Review your schema markup
Structured data is how you tell Google and AIs what type of content you have and how to interpret it. The structured data inspector shows you whether you have it, whether it's correctly implemented, and what's missing.
Generate your llms.txt
Once the above is reviewed, then yes: generate your llms.txt. The automatic generator takes thirty seconds, and then spend another five minutes editing the descriptions to make sure they accurately reflect what you do. That's the part that requires human judgment.
Check if you're already appearing in Google AI Overviews
The citation detection tool tells you whether Google is already mentioning you in its AI responses, or who's appearing in your place for queries related to your industry. That gives you real context on where you actually stand.
If you'd rather have someone walk through all of this for you, instead of running the tools yourself, the Play.Interactive team can review everything and give you a clear picture of where the real priorities are. No endless to-do list, no unnecessary jargon.
And if you want to keep getting this kind of unfiltered analysis every week, subscribe to the newsletter. It lands in your inbox every Monday at 7:45 AM Mexico City time.
Frequently asked questions
The robots.txt tells bots which pages they can and cannot crawl — it's an access instruction. The llms.txt, on the other hand, is a curated summary of what your site does and which pages are most important, designed so that language models have direct context without having to infer it by crawling your entire site. It doesn't block or allow anything; it simply orients.
No. As of this writing, none of the major AI systems have publicly confirmed that an llms.txt is a determining factor in whether they cite a site or not. It's an element that may help, and since it's free and takes minutes to implement, it's worth having — but it doesn't replace the underlying work: content that answers specific questions, properly implemented structured data, and open access for the right crawlers.
All three work together, but with distinct objectives. SEO gives you visibility in traditional search results. GEO — Generative Engine Optimization — optimizes your content so that AI models like ChatGPT or Perplexity recognize you as an authoritative source and cite you. AEO — Answer Engine Optimization — aims to make your site the direct answer to specific questions, like Google AI Overviews. To appear consistently in AI responses, you need all three working together, not just one.
The fastest way is to check your robots.txt manually or use a tool that does it for you. Many sites have ChatGPT's GPTBot, ClaudeBot, or PerplexityBot blocked without anyone at the company knowing — sometimes from inherited configurations, sometimes from security plugins acting automatically. There's a free crawler check at aeon42.com/herramientas that requires no registration.
In that order: first, content that answers specific questions directly and with authority; second, schema markup (structured data) that helps AIs correctly interpret that content; and third, the llms.txt as an additional layer of context. An llms.txt without the first two layers working well contributes very little. The first two without the llms.txt already do most of the work.