AlmaSEO
Always Learning Marketing Assistant
Free tool · No signup
Point it at your site. It reads your sitemap, pulls the real titles and descriptions off your pages, and writes an llms.txt you can publish. No account, no email, no credit card.
llms.txt.yoursite.com/llms.txt — the same place your robots.txt lives.Then the harder part: keeping it true. This file goes stale the moment you rename a page or change a title, and nothing will tell you. AlmaSEO checks whether your llms.txt is present and reachable on every technical audit, alongside the rest of your crawler-facing files.
No. Not yet, and possibly not ever.
llms.txt is a proposal — a convention published at llmstxt.org in 2024, for giving language models a clean, curated map of a site in Markdown instead of making them guess from HTML full of navigation and cookie banners. The idea is sensible. What it isn’t is agreed.
No major AI provider has publicly committed to reading it. Not OpenAI, not Anthropic, not Google, not Perplexity. Plenty of pages will tell you that publishing one will get you cited in ChatGPT or improve your AI visibility. Nobody has shown that, and we’re not going to be the ones to say it just because it would sell better.
So why publish one? Because the cost is close to zero and the downside is nothing. It’s a small static file. It doesn’t affect your Google rankings, it can’t hurt you, and if the convention does get picked up you already have it. That is the honest case: cheap option, unproven payoff. Anyone selling it harder than that is guessing.
The part that is established is duller and more useful: models and crawlers do read your pages, and they read the titles and descriptions you wrote. If those are thin, an llms.txt built from them will be thin too — which is the real signal this tool gives you.
It reads your robots.txt first, because that’s where a sitemap is meant to be declared, and falls back to /sitemap.xml. If your sitemap is an index pointing at other sitemaps, it follows the first one and tells you it did.
For each page it takes your actual <title> and meta description — or the Open Graph equivalents when they’re better. Nothing is invented. If a page has no description, the line just carries the title rather than filling the gap with something made up.
Pages are grouped by the first part of their URL, so /blog/ and /pricing/ become separate sections, with shallower and more important-looking pages first.
llms.txt is meant to be a short guide, not a second sitemap. A thousand-line file is worse than a twenty-line one. If your site is bigger, you’ll be told how many pages were found so you can decide what else belongs.
What it can’t do: it doesn’t run JavaScript, so a site that renders entirely client-side will give thin results. It doesn’t crawl beyond your sitemap. And it can’t judge which of your pages matter most — it guesses from URL depth. Treat the output as a strong first draft, then cut the pages that don’t belong and sharpen the descriptions that read like filler.
The format is Markdown, and deliberately plain: a title, one line saying what the site is, then sections of links with a short description each.
# Acme Plumbing
> Emergency and scheduled plumbing across Greenville, SC. Drain cleaning,
> water heaters, burst pipes, and repiping for homes and businesses.
## Main
- [Emergency plumber in Greenville](https://example.com/emergency): 24-hour callout, typical arrival under 90 minutes.
- [Water heater installation](https://example.com/water-heaters): Tank and tankless, including permits and haul-away.
## Guides
- [Why is my water bill so high?](https://example.com/blog/water-bill): The five leaks that cause most sudden increases.
Two things separate a useful file from a useless one. The description carries the weight — “24-hour callout, typical arrival under 90 minutes” tells a model something; “Learn more about our services” tells it nothing. And it should be short. This file exists to say what matters, and a list of everything says nothing.
The file itself is always llms.txt, with the dot — that’s the filename it has to have to be found. People search for it both ways, and you’ll see “llms txt generator” written without the dot, but there is only one spelling that works on your server.
No. It’s a proposed convention, not a standard, and no major AI provider has publicly confirmed they read it. Nothing breaks if you don’t have one.
No. Google has not said it reads llms.txt, and it plays no part in ranking. If you want to affect what search engines and AI assistants understand about your site, the titles, descriptions and content on your actual pages are what matter — which is exactly what this file is built from.
Nobody has shown that it will. Publishing one is a cheap bet on a convention that may or may not be adopted. Be sceptical of any tool that promises otherwise — that promise cannot currently be supported.
At the root of your domain, so it answers at yoursite.com/llms.txt — the same place your robots.txt lives. A subfolder won’t be found.
No, and they do opposite jobs. robots.txt is an instruction about what crawlers may access, and it is genuinely honoured. llms.txt is a curated summary of what your site contains, offered for models to read. One restricts, the other explains.
No. The generator runs at the edge, stores nothing, and logs nothing. There is no account, no email field, and nothing to unsubscribe from later.
Because llms.txt is meant to be a guide, not a second sitemap. A file listing a thousand URLs is less useful than one listing the twenty pages that actually matter. If your site is larger, the tool tells you how many pages it found so you can add what’s missing by hand.
Partly. It falls back to reading the links on your homepage, which gives a rougher result. Publishing a sitemap.xml is worth doing regardless — search engines use it, and this tool will give you a much better file with one.
Whenever your pages change meaningfully — new sections, renamed URLs, rewritten titles. The file has no expiry and nothing warns you when it drifts out of date, which is the main practical problem with it.
An llms.txt is accurate the day you publish it and quietly wrong three months later. AlmaSEO checks your crawler-facing files on every technical audit, tracks whether AI assistants actually mention you, and tells you which of your pages are worth the attention in the first place.