Chapter 6

A small share of websites got most robot links

A wall covered in sticky notes behind a robot. On the counter stands a small model town: a library with a tall pile of green notes on it, and two houses with small piles. A child looks on, and a man with a clipboard takes notes.

Imagine the shop robot names a toy car, then sticks a note on its answer: read more at this house in town.

Your customers can see the same thing. Many AI answers can show links beside the words, and each link leads to a website.

So I had a computer program count where those links led, in answers about big companies, in 12 languages.

The short version. Most links went to a small share of websites, and thousands of others got a few each. Wikipedia came first in almost every language, with only a small share of links. One Polish question brought links to job interview advice, on pages that seem to have little to do with the company, so open the links you get and read them.

What I did

A computer program sorted the links by website.

On the left, 100 small houses, 18 of them green. An arrow points to 100 sticky notes on the right, 80 of them green.
About 18 of every 100 websites got 80 of every 100 links

Picture it. Each house stands for one in every 100 websites, and each note for one in every 100 links. The 18 green houses got the 80 green notes. The 82 white houses shared the last 20.

At Marta’s: like 18 of every 100 products bringing in 80 of every 100 orders.

Perplexity supplied most of the links, so this pooled picture leans toward it.

Twelve buildings in a row. Eleven are green libraries. The twelfth is a white shop.
Wikipedia came first in 11 of 12 languages, with about 4 to 6 of every 100 links

Picture it. Each building is one language. A library means Wikipedia got the most links there. The white shop is Lithuanian, where another website came first. Below, each square is one link in every 100, and the number counts the green ones.

Polish, Wikipedia’s lowest100 squares, 4 of them green.4
Finnish, Wikipedia’s highest100 squares, 6 of them green.6
Lithuanian, a group of business pages100 squares, 4 of them green.4

At Mikko’s: like one model selling best in 11 of 12 showrooms, at about 4 to 6 of every 100 cars sold.

So coming first still means a small part of the links. A check in English alone would have missed the Lithuanian result. None of this shows that a Wikipedia page helps a company.

In a separate set of Polish answers about 57 companies, one question asked Perplexity to describe a company’s strengths and weaknesses. Some of its answers linked to job sites. I picked four of them after seeing the counts, and looked at where those links went.

100 sticky notes, 94 of them green and 6 white.
About 94 of every 100 links to four job sites led to the same four interview advice pages

Picture it. Each note is one in every 100 links to those sites. Green notes lead to four pages on how to talk about your strengths and weaknesses in a job interview.

At Kadri’s: like picking four suppliers and pulling their papers from all the site files: 94 of every 100 are the same four leaflets, with little to do with those homes.

These job-site links are a tiny part of the Polish set. They show why the page itself is worth opening.

How to check this for your business

Portrait of Joost, the publisher.Joost BakkerPublisher, the NetherlandsMade-up example

Joost asks a robot, in Dutch and in English, which publisher will print a first novel. The answers come with links.

One link goes to a list of publishers. One goes to old news about a firm that has closed. One goes to a writing course with no publisher on the page.

Now Joost knows what the robot puts on display for that question.

Try this on Monday, with your chapter 1 sheet

Take a question from your sheet and ask a robot that shows links, such as Perplexity, in each language your customers use. On a new sheet, write down each website, and tick it each time it comes back.

Open the links with the most ticks and read the pages. Does the page fit the question? Is it right? Is it up to date? Does it name you?

References

Chapter 6. A small share of websites got most robot links

Preprint. Dmitrij Żatuchin (2026). How Large Language Models Source Brand Reputation Across Languages and Markets. arXiv:2606.25787. Data, version 1. 10.5281/zenodo.20829524.

What does AI say about your business?

Put the book into practice. Tell us your company and website to request an AI visibility audit. We’ll reply by email to agree what to check.

What the audit shows

  • Where your company appears in the answers, and next to whom
  • The pages AI cites in your market
  • What to publish first on your own site

Fields marked * are required.

Who is it for?

Your choice about cookies

We use cookies this site needs to work, and others only if you say yes. Choose what you allow, and change your answer whenever you like. Read our privacy policy