llms.txt and Google Analytics on the website

Introduction

After taking care of SEO, sitemap, robots.txt, course pages, FAQ and structured content, I started looking at a new front: how to make the site easier to read by artificial intelligence tools.

That's where the file came in llms.txt.

The idea is simple: if search engines use sitemap.xml e robots.txt To better understand a website, it makes sense to also provide a clear, organized, and straightforward archive for language models, AI assistants, and tools that need to interpret a page's content.

But I didn't want to just publish the file and hope someone accesses it. I wanted to measure.

That's why I configured Google Analytics 4 to know if the llms.txt was being consulted.

What is llms.txt?

O llms.txt is a standardization proposal to place, at the root of a website, a Markdown file with information designed to help language models better understand that website when they need to answer questions or search for context.

In practice, it works as an organized summary of the website.

Instead of an AI needing to navigate through multiple pages, menus, scripts, banners, and HTML structures, the llms.txt offers a cleaner path: it explains what the site is, what the most important pages are and where to find relevant information.

This is important because language models have a practical limitation: they can't put an entire website, with all its details, into the context of a response. The project website itself llms.txt explains that complex HTML, navigation, ads, and JavaScript can make it difficult to extract useful content for LLMs.

Why does this matter to My Robot?

In the case of My Robot Barra da Tijuca, the website is not just a business card.

It explains courses, ages, methodology, Robocopa, My Robot Play, Maker Store, Maker Smart, FAQ, location and pedagogical differences.

In other words: there is a lot of important context there.

When a person asks an AI something like:

“Where are there robotics courses for children in Barra da Tijuca?”

or

“Which school teaches programming and artificial intelligence to teenagers in Rio de Janeiro?”

I want My Robot's content to be as clear as possible to be understood correctly.

O llms.txt It is not a guarantee of ranking, nor does it replace SEO, Google Business Profile, well-written content or advertisements. It is a complementary layer of organization.

Its function is to help AI better understand the website.

llms.txt does not replace sitemap or robots.txt

This point is important.

O sitemap.xml helps search engines find indexable pages.

O robots.txt indicates access rules for robots.

Already the llms.txt offers a curated view of content for language models. The project itself explains that it was designed to coexist with current web standards, complementing sitemap and robots.txt, not replacing these files.

So, in practice, I see it like this:

robots.txt   = orienta acesso de robôs
sitemap.xml  = lista páginas importantes
llms.txt     = explica o site de forma clara para IA

Each one has a different function.

What should the structure of an llms.txt be like?

The recommended format is Markdown.

The file should preferably be located in the root of the site, in /llms.txt, and must have at least one main title with the name of the project or website. It may also have a summary, additional explanations, and lists of links organized by sections.

A simple example would be:

# My Robot Barra da Tijuca

> Escola de robótica, programação e tecnologia educacional para crianças e adolescentes na Barra da Tijuca, Rio de Janeiro.

## Páginas principais
- [Cursos](https://www.exemplo.com/cursos.html): cursos de robótica, programação, IA e tecnologia.
- [FAQ](https://www.exemplo.com/faq.html): principais dúvidas de famílias e responsáveis.
- [Contato](https://www.exemplo.com/contato.html): endereço, WhatsApp e canais de atendimento.

The ideal is to use objective language, well-described links and avoid ambiguous terms. The project itself recommends clear language, informative descriptions in links and testing with language models to see if they can respond well to the site's content.

The problem: Google Analytics doesn't measure llms.txt automatically

After I published the file, a practical question came up:

“How do I know if someone is accessing the llms.txt?”

The first attempt was to look at GA4 in real time. But there is a technical detail: a file .txt it is not an HTML page.

It does not load the Google Analytics script.

He doesn't have <head>.

It does not run JavaScript.

So if I simply publish the file as plain text, GA4 probably won't measure access automatically.

That's why I needed to create a server-side measurement.

The solution: send an event to GA4 via Measurement Protocol

To resolve this, I used Google Analytics 4's Measurement Protocol.

The Measurement Protocol is a way of sending events directly to Google Analytics servers, via HTTP requests. Google's own documentation explains that it allows you to send data to Analytics differently than gtag, Google Tag Manager or Firebase, requiring events to be scheduled manually.

In my case, the logic looked like this:

1. Alguém acessa /llms.txt
2. O servidor entrega o arquivo normalmente
3. Ao mesmo tempo, o servidor envia um evento para o GA4
4. O evento aparece no Analytics como llms_txt_access

The event I created was:

llms_txt_access

With this, I can know if the URL is being consulted without relying on JavaScript in the browser.

How I set up measurement in GA4

The process had two parts: one in Google Analytics and another in the website code.

1. I got Measurement ID

In GA4, I went to:

Administrador > Coleta e modificação de dados > Fluxos de dados

Then I selected the website's web flow and copied the metrics ID, which has this format:

G-XXXXXXXXXX

This value was saved as an environment variable:

GA4_MEASUREMENT_ID

2. I created the Measurement Protocol Secret API

Still in the web flow, I accessed:

Chaves secretas da API Measurement Protocol

Then I clicked on create a new key.

This key was saved as:

GA4_API_SECRET

There is an important precaution here: this key must not be exposed in HTML, public JavaScript or the GitHub repository.

How the event is sent

The server-side implementation sends a POST to the GA4 Measurement Protocol endpoint.

The payload used follows this idea:

{
  "client_id": "id_anonimo_gerado_no_servidor",
  "events": [
    {
      "name": "llms_txt_access",
      "params": {
        "file_path": "/llms.txt",
        "file_url": "https://www.myrobotbarra.com.br/llms.txt",
        "content_type": "text/plain",
        "source_type": "server",
        "page_location": "https://www.myrobotbarra.com.br/llms.txt",
        "engagement_time_msec": 100,
        "session_id": 1234567890
      }
    }
  ]
}

I also included session_id e engagement_time_msec, because these parameters help the event appear correctly in reports such as Real Time.

What not to send to Google Analytics

This point is fundamental.

I do not send name, telephone number, email, IP, WhatsApp, student data, address or any personal data at this event.

The goal is just to know that the file was accessed.

So the event only measures something technical:

Alguém ou algum robô acessou /llms.txt

It doesn't measure who the person was.

Does not identify the visitor.

It doesn't turn this into a lead.

How did I validate that it worked

After implementation, I did the simplest test:

1. Acessei /llms.txt no navegador.
2. Voltei ao Google Analytics.
3. Entrei em Relatórios > Tempo real.
4. Verifiquei se /llms.txt apareceu na tabela de páginas.

And it worked.

GA4 now shows real-time file access.

After that, you can also consult:

Relatórios > Engajamento > Eventos

And search for the event:

llms_txt_access

Should this event be conversion?

In my view, no.

Access to llms.txt it's not a lead.

It's not a click on WhatsApp.

It is not a completed form.

It is not a trial class schedule.

It is a technical event.

It is used to understand whether tools, bots, crawlers or users are consulting the file. Therefore, I do not recommend marking llms_txt_access as conversion in Google Ads.

The important conversions continue to be:

clique no WhatsApp
envio de formulário
clique em telefone
agendamento de aula experimental

What I learned from this

The main learning is that technical SEO and artificial intelligence are getting closer and closer.

Before, the concern was just:

O Google consegue encontrar minhas páginas?

Now the question also begins to be:

As ferramentas de IA conseguem entender corretamente o meu site?

O llms.txt enters precisely at this point.

It doesn't solve everything, it doesn't replace good content and it doesn't guarantee automatic visibility. But it helps organize website information more clearly for language models.

And measuring this access in GA4 helps remove the doubt from the realm of supposition.

Instead of just publishing the file and waiting, I can now track whether it is being queried.

Conclusion

Create a llms.txt It is a simple but strategic initiative.

For an educational website like My Robot Barra da Tijuca, it helps to better explain the proposal, courses, methodology, main pages and the most important paths for those looking for information about robotics, programming and educational technology.

But publishing the file is only half the work.

The other half is measuring.

By configuring a server-side event in Google Analytics 4, I was able to know if the file is being accessed and validate that the URL appears correctly in real-time reports.

In practice, this is the type of technical adjustment that does not appear to the average visitor, but improves the website's digital organization and better prepares the brand's presence for an environment where search, AI and structured content increasingly go hand in hand.

Next step

Now that the event is up and running, I would track it weekly:

Relatórios > Engajamento > Eventos > llms_txt_access

And would keep llms.txt updated whenever new important pages are created, such as courses, events, Robocopa, FAQ, blog articles and conversion pages.

What does this have to do with learning programming?

Behind a configuration like this there is something bigger than “tinkering with a tool”: there is programming logic, documentation reading, data organization, integration between systems, care for privacy and the ability to transform a real need into a functional solution.

These are exactly the types of concepts that bring young people closer to the professional world of technology. When a student understands how an event is sent to an analysis tool, how an API receives data or how a system needs to handle information securely, they begin to realize that programming is not just writing code: it is building bridges between problem, logic and result.

Related Products

Products to understand programming beyond the screen

Although the article is technical, it talks about logic, events, data and integrations. These products help young people understand how code and electronics can turn into real solutions.

Board compatible with Arduino Uno R3

An affordable foundation for connecting programming, inputs, outputs, and physics experiments.

View in Maker Store

Arduino Maker Store Kit

Suitable for exploring logic, sensors, automation and first physical computing projects.

View in Maker Store

Maker Connect 52 in 1 Kit

Helps connect programming, assembly, and problem solving into practical robotics projects.

View in Maker Store

Affiliate links: by purchasing through these links, you support My Robot Barra da Tijuca.

Sources consulted

• llms.txt — official proposal
• Google Analytics 4 — Measurement Protocol
• Google Analytics 4 — Sending events
• Google Analytics 4 — Measurement Protocol Reference
• Google Analytics Policy on Personally Identifiable Information