Chapter 10

When language and place both changed, the names changed

A toy shop. A robot stands behind a counter. Two children stand at the counter, one at each end. One has a map of a wide area, and the robot's bubble on that side shows one big shop. The other has a map of a small place with a pin in it, and the bubble on that side shows the big shop and a small one.

Imagine two children in a toy shop. One asks a robot, in English, for the top toy shops in Europe. The other asks in Lithuanian about toy shops in Lithuania. A shop known only in Lithuania comes up for the second child.

Your customers can ask in any language, about any place. A local name can show up in one answer and be missing from another.

So I asked robots about well-known companies from eleven European countries, in twelve languages.

The short version. Local names came up far more when the question used the language of the company’s head office and named the area around it. Those two things changed together, so I cannot say which one mattered. When you test, change one thing at a time.

What I did

The robots could search the web.

Local names came up far more in the home version

I used two questions that did not name the company, asked in English and in the company’s home language. Local names came up far more in the home version. For big global names, the gap was much smaller.

Named means only that the answer said the name, in praise or in passing.

For a Lithuanian power company, the English questions named Europe or “International”. The Lithuanian versions named the Baltic states or Lithuania. Language and place changed together.

Four boxes in two rows and two columns. A speech bubble marks each row, and a map marks each column. The top left and bottom right boxes are green, with a question card in each. The other two boxes are grey and empty.
I asked 2 of the 4 mixes, so I cannot split language from place

Picture it. Each box is one mix of language and place: English on top, home language below, a wide area on the left, home on the right. Green boxes are mixes I asked. Grey boxes stayed empty.

At Javier’s: like trying 2 of 4 mixes, old flour in the old oven and new flour in the new oven. If the pizza tastes different, nobody can say why.

A tone program saw gaps

A second program gave each answer a mark from minus one to plus one. I compared the English and home-language marks for each company.

65 folders in rows of ten. 29 are green and 36 are white.
The program saw a gap above 0.15 on 29 of 65 companies

Picture it. Each folder is one company. On a green folder, the marks for its two sets of answers sat more than 0.15 apart. On a white folder, they sat closer.

At Daniel’s: like a program reading 65 client files, each in German and in English, and giving the German copy the higher mark by more than a set margin on 29. The English copies run longer, and the program reads only the start of each.

Do not trust such a mark on its own. The English answers were the longest, and I have no count of what length does to the mark. So read the answers yourself.

Repeats differed more by robot than by language

In another part of the study, I asked a few questions five times each, one company at a time.

100 parcels. 87 are green and 13 are white.
In 87 of every 100 sets of repeats, the meaning stayed very close

Picture it. Each parcel is one set: five answers from one robot, in one language, to one question. Green means a computer found the five very close in meaning.

At Marta’s: like 100 customers each ordering the same thing five times, and 87 of them getting five boxes that look almost the same.

How to check this for your business

Portrait of Giulia, the accountant.Giulia RomanoAccountant, ItalyMade-up example

Giulia asks a robot, in English, for the top accountants in Europe. The answer lists big firms. Then Giulia asks in Italian about accountants in Milan, and the list changes.

Was that the language or the place? Giulia changed both, so there is no way to tell yet. Next comes the mix Giulia skipped: English, and Milan.

Try this on Monday: fill the empty box

Take one question from your chapter 1 sheet. On one day, with the same town, ask it ten times for each robot in your sheet’s language, then again in English, in a new chat each time. If your sheet is in English, use another language your customers speak. A translation website can help.

Compare the two sets of ticks, and read what each answer says about you. You changed one thing, so this is cleaner than mine.

References

Chapter 10. When language and place both changed, the names changed

Preprint. Dmitrij Żatuchin (2026). The Language Blind Spot: How Query Language and Brand Recognition Tier Shape AI-Constructed Brand Reputation Across Twelve European Languages. arXiv:2606.23165. Data. 10.5281/zenodo.20794390.

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