Looking for an llms.txt generator? You probably don’t need one

Before you install another plugin or pay someone to create an llms.txt file, ask what you expect it to do.

If the answer is “get my website into Google’s AI answers,” you can stop. Google explicitly says it ignores these files for search visibility and rankings, including its generative AI features. An llms.txt generator cannot change that. Google’s AI optimization guide

There are legitimate reasons to publish the file, especially when people use AI agents to work with your software documentation. There are also convenient ways to generate it.

But a website owner looking for more customers deserves a clearer explanation than “AI needs this.”

I looked at the promises on a generator’s website, the proposal behind the file, Google’s documentation and research into whether anything actually requests these files. My advice: establish a use for the output before choosing a tool to produce it.

First, this is a website file

The correct name is llms.txt. The “llms” refers to large language models; the extension is .txt. An llms.txt generator creates a file about your website.

The proposal describes a Markdown document containing a site or project name, a summary and organized links to useful resources. Markdown is ordinary text with simple formatting: a # introduces a heading, for example.

Imagine a software company with hundreds of documentation pages. A short file could point an agent toward the installation guide, authentication instructions and API reference.

That is an understandable use. It gives software somewhere useful to start when someone directs it to the documentation.

It does not, by itself, explain why an AI search service should recommend that company to a potential customer.

Read what the generator is promising

The LLMrefs generator pages usually describe free tools that crawl a website, uses AI to write page descriptions and produces an editable file. Those are concrete features: it can save you some copying, organizing and writing.

The same page goes further. It calls the file “essential for GEO/AEO” and tells users that “AI tools will automatically find and use this file.” It also presents better recommendations and a competitive advantage as benefits.

Those claims deserve separate scrutiny.

Producing a file is something the generator controls. Whether an outside service discovers it, uses its contents and changes its recommendations is outside the generator’s control.

Google has already answered the main SEO question

Google’s AI optimization guide specifically addresses llms.txt under its mythbusting advice.

It says maintaining these files will “neither harm nor help your site’s visibility or rankings” in Google Search.

Google also explains that it may discover and crawl a file without giving the format special treatment. Seeing Googlebot request your llms.txt therefore would not contradict its guidance.

For a business hoping to appear in AI Overviews or AI Mode, this removes the main reason to treat an llms.txt generator as an SEO requirement.

That statement applies to Google Search. It should not be stretched into a claim that every AI product ignores the file.

The biggest number needs its small print

Ahrefs studied 137,210 domains using its Web Analytics product and examined requests to /llms.txt during May 2026.

Around 38,000 domains had a file that passed its checks. Of those files, 97% received no requests in the measured period.

That is a serious challenge to the idea that publishing one automatically attracts AI readers.

However, it was a study of an Ahrefs customer population over one month. It measured requests, not a controlled change in rankings or citations. It also predates the proposal’s August 2026 revision.

The finding supports skepticism about automatic discovery. It cannot establish that every file is useless, or that no agent will ever use one.

The distinction also works in the other direction: a recorded request does not prove the contents influenced an answer. Downloading a file, following its links and recommending a business are different outcomes.

There is a real use case, and it may already be covered

Suppose your customers use coding agents to integrate your product. Helping those agents locate the correct documentation can be useful.

You can see an actual implementation in Anthropic’s developer documentation index, which lists documentation resources.

Publishing such an index demonstrates that a company offers the file. It does not demonstrate that its AI products routinely use other websites’ files to select search recommendations.

Documentation platform Mintlify automatically generates and hosts llms.txt, keeps the generated file current and advertises it through response headers. If your documentation already uses that feature, you may have no reason to find a separate generator.

There has also been a meaningful update to the proposal. Its August 2026 revision adds guidance for linking a page to its Markdown version and the relevant llms.txt file. The change addresses discoverability: how an agent can find the resources without guessing.

That makes checking the implementation more useful than checking a box marked “file exists.”

For a documentation team, the practical questions are which agents customers use, how those agents find your resources, and whether they can complete the intended task accurately.

If you actually need a file, you can write one

For a small documentation site, a text editor is enough. Here is an illustrative starter:

# Example Booking API

> Documentation for developers integrating the Example Booking API.

## Start here

- [Setup](https://example.com/docs/setup.md): Create a test integration.
- [Authentication](https://example.com/docs/authentication.md): Authenticate requests.

## Reference

- [Reservations](https://example.com/docs/reservations.md): Create and update bookings.
- [Errors](https://example.com/docs/errors.md): Interpret error responses.

The title, summary and grouped links follow the basic structure described in the llms.txt specification.

Replace the example URLs and descriptions with your own. Only link to Markdown versions that actually exist; adding .md to a URL does not create a page.

Save the document as llms.txt and have your website serve it at the intended public location, such as /llms.txt or /docs/llms.txt. Check that the URL returns the actual text and that its links work. Follow the proposal’s discovery guidance if you are implementing it for compatible agents.

A generator becomes useful when it saves meaningful work maintaining a larger collection. For a small file, you can judge whether selecting and describing a few links yourself is quicker than setting up another tool.

Either way, someone needs to check the result.