llms.txt Generator
Google Search ignores this file — its own documentation says so, and no AI provider publishes support for reading it. So here is the honest version of this tool: it builds a spec-valid llms.txt, it validates it properly, and the panel beside it tells you who the file is actually for. Decide from there.
Your file
Your llms.txt
Validator
yourdomain.com/llms.txt — containing an H1 with your site name, a blockquote summary, and H2 sections of curated markdown links. It is a human-readable index of your best content. Google Search does not use it, no AI provider publishes support for it, and it is still cheap enough to be worth having if you treat it as documentation rather than as SEO.How to use the generator
- Name the site and write the summary The name becomes the H1 and is the one element the llms.txt spec actually requires. The summary becomes the blockquote directly under it, and it is the most important line in the file — it is the sentence a reader uses to decide whether the rest is relevant.
- Add your sections One section per kind of content: guides, tools, services, docs. Each line inside a section is one link. Use
Title | URL | note, or paste a markdown link —[Title](URL): note— and the tool reads either. - Mark anything skippable as Optional Tick the Optional box on a section and its heading renders as
## Optional. That is the spec's convention for links a reader or agent can ignore when it needs a shorter version of the file. Use it for background reading, not for your main pages. - Read the validator before you copy It is the part no other generator has. Fix the red items, consider the amber ones, then copy or download. The file goes at the site root, nowhere else.
What the validator checks
The file has no official linter, so these rules come from the spec plus the failures we keep finding in llms.txt files in the wild — mostly relative URLs copied out of a CMS, and sitemap dumps with four hundred links in them.
| Check | Level | Why it matters |
|---|---|---|
| URL is not absolute | Error | A relative path like /guides/seo/ is meaningless to anything reading the file away from your site. Every URL needs the scheme and host. |
| Entry has no title | Error | The spec's link format requires a name. A bare URL tells a reader nothing about what is behind it, which defeats the point of a curated index. |
| Characters that break the link syntax | Error | A ] in a title or a space or ) in a URL breaks [name](url) as markdown. The file stops parsing cleanly and nobody tells you. |
| Duplicate URL | Error | The same page listed twice wastes the reader's attention and usually means two sections overlap. |
| Near-duplicate URL | Warning | Two entries differing only by a trailing slash are almost always the same page. Pick whichever one your canonical tag points at. |
| Link points off your domain | Warning | Fine if deliberate. Usually it is a URL pasted from staging, or a host typo. |
| Section has no links | Warning | An empty H2 is noise. Delete the section or fill it. |
| Two sections share a heading | Warning | Merge them. Two identical H2s read as a mistake, and two ## Optional sections defeat the convention. |
| More than 10 links in one section | Warning | This is a curated index, not an XML sitemap. Past ten, split the section or cut the weakest entries. |
| File over 10,000 characters or 50 links | Warning | The whole value of the file is that it is short enough to be read end to end. Past this, it is a sitemap with extra steps and nothing will read all of it. |
Where the file goes
At the root of the domain, as a plain text file, reachable at https://yourdomain.com/llms.txt. Three things go wrong here more than anything else:
- It is a file, not a redirect. Do not 301
/llms.txtto a page, a docs URL or an S3 bucket on another host. Anything reading it expects the content at that path. - It is not a page. Do not publish it as a WordPress post or an HTML template with your header and footer wrapped around it. Serve the raw markdown, ideally as
text/plainortext/markdown. If your server hands backtext/html, the content is still readable, but a strict reader may refuse it. - It lives at the root, not in a folder.
/docs/llms.txtis not the convention and will not be found by anything looking for the file. Subdomains each need their own.
There is no submission step, no Search Console equivalent and nothing to ping. Upload the file and you are done. If you want a second, longer file with the full text of your docs concatenated, that convention is llms-full.txt — the same idea at a size no one should paste into a context window by hand.
llms.txt is not robots.txt, and the difference is the whole point
These two get confused constantly, and the confusion runs in the expensive direction: people publish an llms.txt believing they have said something about AI crawlers, when the file that governs crawlers is sitting untouched a few bytes away.
| llms.txt | robots.txt | |
|---|---|---|
| What it is | A convention proposed in September 2024. No standards body, no adoption commitment from any provider. | A standard from 1994, formalised as RFC 9309 in 2022, honoured by every major crawler. |
| What it does | Suggests what to read. Invites. | States what may be fetched. Restricts. |
| Who reads it | Nobody who publishes it. Tools you hand it to. | Googlebot, Bingbot, GPTBot, ClaudeBot, PerplexityBot, CCBot and the rest — all of them check it. |
| Effect on AI crawlers | None. It cannot grant or deny access. | Direct. A Disallow for GPTBot or ClaudeBot is the actual lever. |
| Effect on Google Search | None. Google's guidance says Search does not use it. | Substantial. A bad rule here can remove your site from search. |
If the reason you came here was to do something about AI crawlers, the file you want is the other one. Our robots.txt generator builds it, and the robots.txt guide explains which AI user-agent does what before you start blocking them — worth reading first, because the rules are easy to get subtly wrong. When we read the robots.txt of 99 Indian companies in September 2026, the most interesting failure was a news publisher whose file named six AI crawlers and then permitted every one of them, because the user-agent lines were stacked with no rule between them. It looked like a blocklist. It was an invitation.
What this tool cannot do
It runs entirely in your browser and makes no network requests at all. That is deliberate — nothing you type is uploaded — but it rules out a whole category of things the crawler-based generators claim to do, and one thing nobody can do.
- It cannot crawl your site. It will not discover your URLs, read your sitemap or pick your best pages for you. You choose what goes in the file, which is the correct way round: a machine guessing your twenty most important pages produces exactly the bloated, uncurated file the validator warns about. If you want an automated read of your site, that is an AI search readiness check, not this.
- It cannot verify a single URL. No status codes, no redirect chains, no check that the page exists or that the title you typed matches what is on it. A perfectly valid llms.txt full of 404s validates clean here. Click through your own links once before you publish.
- It cannot see your robots.txt. If a URL in your file is disallowed to crawlers, the tool has no way of knowing, and you will have published an index pointing at pages nothing is allowed to fetch.
- It cannot upload anything. Copy or download the file, then put it at the site root yourself — that step needs FTP, your host's file manager or a deploy.
- It cannot make anything read the file. Neither can any other generator, however it is marketed. Publishing an llms.txt is publishing a document and hoping it is useful. Treat it as documentation and it is a reasonable twenty minutes; treat it as an AI-visibility tactic and you have been sold something.
One more: the character and link thresholds in the validator are our editorial judgement, not spec requirements. The spec sets no size limit. We flag at 10 links per section and 10,000 characters overall because past that the file stops being a curated index — but if you have a genuine reason to go longer, go longer. The warning is a nudge.
Where to go next
If llms.txt brought you here because you are trying to appear in AI answers, the file is not the lever — the AI search guide covers what actually is, and Google AI Overviews explains how that surface picks its sources. On the technical side, structured data and the fundamentals in the complete SEO guide do more for machine readability than any new text file will. The rest of the free tools cover the neighbouring jobs.
Frequently asked questions
Does llms.txt help me rank in Google or appear in AI Overviews?
No, and Google says so in writing. Its guide to optimising for generative AI features, last updated 10 July 2026, states: “You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn't use them.” The separate AI features documentation, last updated 10 December 2025, says there are no additional technical requirements for AI Overviews or AI Mode beyond the ordinary Search requirements, and never mentions llms.txt. Publishing the file will not hurt you either — it is simply not part of how Google works.
Does any AI tool actually read llms.txt?
None that publish support for it. We checked on 29 September 2026: there is no documented llms.txt support from OpenAI, Anthropic, Google or Perplexity. Anthropic publishes an llms.txt of its own documentation, which is not the same as saying its crawlers consume one. Google's John Mueller said in June 2025 that no AI system was using the file, and nothing has superseded that. What does read it: tools and agents you point at the URL yourself, which is a real use case as long as you are honest that it is a file you hand over rather than one that gets found.
llms.txt or robots.txt — which one controls AI crawlers?
robots.txt, without qualification. It is a standard from 1994, formalised as RFC 9309, and every significant crawler including GPTBot, ClaudeBot, PerplexityBot and CCBot checks it before fetching. llms.txt cannot grant or deny access to anything; it is a reading suggestion. If your goal is to allow or block AI crawlers, use our robots.txt generator and read the robots.txt guide first, because the grouping rules are easy to get wrong in a way that silently permits everything you meant to block.
How many links should an llms.txt file have?
Few enough that someone reads all of them. The spec sets no limit, so this is judgement: we warn past ten links in a single section and past fifty or 10,000 characters overall. The value of the file is curation — if it lists every URL on the site, it is a sitemap in a worse format, and anything reading it will truncate long before the end. If you genuinely need the full text of your documentation in one file, that is the llms-full.txt convention instead, and it is a different job.
Is anything I type into this tool sent to you?
No. The file is built in your browser as you type, the validator runs in your browser, and the download button uses a client-side Blob — no upload, no API call, nothing stored, nothing remembered between visits. Refresh the page and it is gone. The tool works with the network disconnected once the page has loaded, which you are welcome to verify.
A text file will not get you cited by an AI answer.
Send us your site. We will tell you which of your pages AI answer engines can actually use, what is blocking the rest, and which of the popular AI-search tactics are worth your time.