How does the website get into the response

We are talking about one thing: how to make a website readable for AI-powered search and language models. GEO, llms.txt, structured content, and the difference between when a page is merely indexed and when it is cited.

These are not studio news. We don't write about ourselves here.

How to get a link to your business in Google Maps

You can copy a business link from Google Maps in half a minute. Steps on a computer and a phone, an embedded map for your website, and the difference between a short and a long link.

Why are we writing about it

This blog is written by a studio that builds websites and applications. We don't select topics according to an editorial plan. They're questions that arose during actual development and needed to be resolved before launching the site.

We therefore write about what we had to research and verify ourselves. If it turns out that something works differently than commonly claimed, we will write about it, including what remains unclear.

And we're publishing this mainly because almost nobody writes about it in Czech. What's available is mostly in English and written for large websites. A company with a website of just a few pages will get little value from it.

What are we focusing on here

Circles to which it keeps returning. Each of them addresses a different part of the same problem: making the web meaningful to those who don't read it with their eyes.

GEO Optimization for answers, not for positions.: The search engine rewards you with a spot in the list of links, the model by using you within the response it writes itself. These are two different disciplines, and they can work against each other. GEO focuses on the latter: making it possible to take a piece of text from a page that makes sense even when removed from its context.
llms.txt Sitemap written for model.: A text file at the root of the site that says which pages are worth attention and what each one is for. It isn't required, and nothing stops working without it. It earns its keep once a site has more pages than a model will crawl through, and you want to say which ones it should start with.
Structured data Description of what is on the page.: Tags that tell machines this is the price, this is the service, and this is the opening hours, instead of letting it guess from formatting. You're not enforcing anything, but you're removing areas where the model is uncertain. We'll discuss which types make an impact for a small business and which are added merely out of habit.
Content structure and writing The page layout and a paragraph that stands on its own.: How to structure a page so that it contains one topic and one answer. Headings work as addresses that the model uses to navigate, and the text beneath them should make sense even without what precedes it. Most websites are written the opposite way: as continuous narratives from which nothing can be extracted.
Measurement What can be seen from it and what cannot.: The citation won't appear as a clickable link, so from the outside it looks like nothing is happening. You can track indirect traces: views with no clicks, and whether more people are searching for your name directly. We also write about things that can't be measured and that you shouldn't trust in reports.

Questions that arise around it

What is GEO and how does it differ from SEO?

SEO focuses on getting the page to appear as high as possible in the list of links. GEO focuses on getting the model to use the page within its own generated response. These don't automatically go together: a page can rank first in search results and still not appear at all in responses. Likewise, it can be cited in responses without ever making it into the top ten results.

Do we need to have llms.txt?

You don't have to. It's not mandatory, and nothing will break if the website doesn't have it. It's a sitemap written for the model: which pages are important and what each of them is for. It makes sense when a website has more pages than the model will bother to crawl, and you want to indicate which ones it should start with.

Will structured data help the model to cite us?

Structured data tells the machine what the thing on the page is (that this is a price, this is a service, and this is opening hours), not how it looks. You won't force a citation this way; no markup can do that. But you eliminate the guesswork that most often prevents a citation: the model doesn't have to infer what you meant by that piece of text.

How do we know that our website is being cited?

Hard to tell, and that's the unpleasant part. Citations don't show up as clicks, so in your traffic analytics you see nothing and it appears as if nothing is happening. You can only track indirect traces: whether views are increasing without any clicks afterward, and whether more people are searching for your name directly. Alternatively, you can directly ask models a question from your field and see whose names they mention.

Do you also do it to order?

Yes, but not as a separate service that you order separately. It's part of what we build. We make the website or application right away so that it can also be read by machines. There are two ways to get started.

There are no additional articles here yet.

Every text is published in three languages

The topic has readers in Czech, English, and Russian. Articles therefore do not appear first in one language and then get translated into the others. They are available in all three languages from the start.

Czech English Russian

When someone has to do it for you

Sitely builds custom websites and applications. This is one of two ways to get started with us.

New website or application

Tell us what the website or app should do. We'll design it, you approve the design, and only then will it be built.

Pop out project

You have a website and want it different

Send the address of your current website. We'll review it and let you know what can be done with it.

Send website address