llms.txt generator and validator

How to use it

1. 01Write the site's official name and a one- or two-sentence summary that says what it is, without sales adjectives.
2. 02List the pages that answer your audience's central questions, one per line, as Title | URL | description. Group them with lines starting with ##.
3. 03Click generate. Read the checks: an error breaks the format; a warning or a tip is editorial advice.
4. 04Copy the text or download the file. Publish it as /llms.txt at the domain root, served as plain text, and open the address in a browser to confirm.
5. 05Review the file when pages change. A broken link or a stale description is worse than no file.

What does Google say about llms.txt?

Google says Search does not need it. In the guide "Optimizing your website for generative AI features on Google Search", last updated on 10 July 2026, the company states that you do not need to create machine-readable files, AI text files, markup or Markdown to appear in Google Search, and that Google Search does not use them. According to the same guide, creating and maintaining such files for other services is fine, and the effect on Search visibility is nil in both directions. The page on AI features, updated on 10 December 2025, repeats that there are no additional requirements or special optimizations to appear in AI Overviews and AI Mode.

That position covers Google Search. It does not describe what ChatGPT, Claude, Perplexity, Copilot or coding agents do, and it does not prove that no agent anywhere reads the file. It shows only that, for Google Search, the file does not count.

What does the format propose, according to its author?

The proposal is by Jeremy Howard, published on 3 September 2024. The llmstxt.org page, which shows version 2 and a modification date of 10 August 2026, describes a Markdown file at the site root (or at a subpath, which applies to the pages beneath it) with five parts, in this order:

By convention, an H2 section named "Optional" holds secondary links that an agent may skip when it needs a shorter context. The page itself calls the overview informal.

The llms.txt of the Model Context Protocol documentation site, opened on 14 October 2026, is a simple real example: it starts with the title "Model Context Protocol" and continues with a list of links, some with descriptions, and no summary block. It shows that only the H1 is required.

Who reads the file?

Less is known than many people assume. What is public: the proposal says the file serves agents at query time, and technical documentation sites publish it. The SE Ranking study of 7 November 2025 analysed nearly 300,000 domains, 10.13% of which had the file, and found no correlation between having it and citation frequency; the authors say the result depends on the model and dataset. It is a vendor study, correlational, and it does not measure use by coding agents. An honest summary: there is no public evidence that the file increases citations, and no sign that it hurts.

For the full discussion in Portuguese, with a table of what is and is not known, read llms.txt funciona?. For the decision that really changes crawler access, see the guide on robots.txt for AI crawlers (in Portuguese): robots.txt para robôs de IA.

What does the validator check?

The validator separates what belongs to the format from editorial advice. The source of the format is the llmstxt.org page; the editorial advice is the tool author's.

The tool does not visit the addresses. It cannot tell whether a page responds, is blocked in robots.txt or has noindex. Those tests are yours.

Filled-in example

Click "Load example". The tool uses a fictional plumbing company and produces a file like this:

# Model Plumbing

> Fictional company offering residential and commercial plumbing services in two cities. This text is a filled-in example of the generator.

Open Monday to Saturday. Prices depend on the scope and are listed on each service page.

## Services
- [Sewer unclogging](https://example.com/unclogging/): scope, response times and what the price includes.
- [Leak detection](https://example.com/leaks/): inspection method and when it is indicated.

## Company information
- [About us](https://example.com/about/): licensed technicians, registration and service areas.
- [Frequently asked questions](https://example.com/faq/): response times, payment and warranty.

## Optional
- [Privacy policy](https://example.com/privacy/): how contact data is handled.

Names, address and descriptions are fictional. When you replace the fields with your own, keep descriptions factual: what the page answers, not why it is good.

What about llms-full.txt?

Some documentation sites also publish a file with the full text of their pages, usually called llms-full.txt. That name does not appear on the proposal page opened on 14 October 2026. Treat it as a convention of whoever publishes it, not part of the format, and do not generate one without a concrete reason, such as an audience of developer agents that consume your documentation.

When it is not worth the effort

If your time is short, other actions have more support in official sources: making sure important pages are indexed and can show a snippet, allowing search crawlers, fixing outdated facts on profiles and directories, and writing the answer at the top of each page. llms.txt comes after, if there is room to maintain it. If nobody will update it, do not create it. This site publishes its own /llms.txt, generated automatically from its pages, as a low-cost experiment with no expected effect.

Limits of the tool

• It visits no address. It does not check links, HTTP status, robots.txt or noindex.
• It proves no effect. A file with no validator errors is well formed; that does not mean anyone reads it or that the brand will be cited.
• The promotional-language check works by words. It can raise false positives and miss exaggeration written another way. Reread the summary.
• Nothing is saved. If you reload the page, the fields are lost. Copy or download before leaving.
• The download uses UTF-8 and LF line endings, with no byte-order mark.

Frequently asked questions

Next step

If you plan to roll the file out across several sites, the JSON-LD and llms.txt template pack by Camillo Dantas bundles llms.txt, robots.txt and structured-data templates with a validation guide: see the pack at GEO SEO Academy (page in Portuguese). To decide first whether the file is worth the effort in your case, go back to llms.txt funciona? (in Portuguese).

Last reviewed: 14 October 2026. Google's position and the use of the format may change; check the sources before deciding. Tool of the LLM Seeding method, authorship declared by Camillo Dantas.