How AI Search Finds a Page: The Answer First, Then DNS

Blog 3 min read

A model cites an answer that sits in public HTML. The inside of a PDF, a locked panel and broken DNS are not a citation.

ChatGPT or a Google AI summary does not “rank” a page. It cites text it can see, on a public address, that answers the question in the first paragraph. Writing a keyword ten times does not buy that citation. A locked client panel, a PDF behind a login and DNS that does not resolve are invisible.

EuroVDC’s machine-readable profile is llms.txt. That file describes the services and where live prices are. It does not replace the page a visitor reads. The thing that gets cited is still HTML.

Four conditions for being seen

  1. The address resolves. If the zone is missing or the name has expired, there is no page to read. When the nameserver is ns1.eurovdc.eu, the records are there.
  2. The page is public. Behind a login wall is not a citation.
  3. The first paragraph answers the question. “In this article we will discuss” is not an answer. The answer is the date, the number and the rule, written there.
  4. The address stays. A new slug every week leaves the citation pointing at nothing. A stable address is worth more than a campaign URL.

What speed does

A slow page does not guarantee a citation, and a fast page does not either. A page the reader cannot open is a page a model also times out on. We do not invent a millisecond. The time of a page in Sofia is what the looking glass and your own city say.

What this is not

Adding llms.txt does not make you the first line of a chat. The file exists so a model can describe the service accurately. An invented price or an invented delay does not belong in that file or on this site. The live price is on the product page.

What the first paragraph has to do

An opening that does not get cited: “In this article we will discuss domain names.” An opening that can: “At 18:00 UTC on 7 October 2026 ICANN publishes the application list. Registration does not open that evening.” The second one puts the date, the fact and the rule in the first sentences.

Four languages are four pages. A Turkish answer is not a citation for a German question. Each language needs its own address and hreflang. llms.txt is the service profile. It does not replace the first paragraph a visitor reads. Do not write a price on the page that is not in the profile. The live price is on the product page.

Frequently asked questions

If I close robots.txt, will AI stop seeing the page?

A crawler that follows the rule will not. Not every tool follows it. If you want it hidden, do not publish the page. A robots line is not a secret.

Is FAQ schema required?

No. Schema is a copy of the answer on the page. Writing an answer into schema that is not on the page gives the citation a false sentence.

Does each language need its own page?

Yes. A Turkish answer is not a citation for a German question. Each language answers its own question in its own first paragraph.

Did EuroVDC sell this article to a model?

No. The page is public. A model can read a public page. That is not a purchased ad.

ai search llms.txt dns citation

EuroVDC

Find Your Perfect Domain Name

500+ extensions · Instant activation · Free DNS management

Search for a Domain

Did you find this content useful?

– People found it useful

Share on Social Media

How AI Search Finds a Page: The Answer First, Then DNS