AI assistants may recommend different brands in other languages

AI recommendations in different languages can name different businesses. Ask an assistant to recommend a business like yours in English, then ask the same question in another language your customers use, and the lists that come back may not match. Research covered by Search Engine Journal found exactly that.

What the research found

Search Engine Journal covered a study that looked at 1,650 responses and found AI assistants recommended different brands depending on the language the question was asked in. The item is dated October 2026.

The finding is about which brands are named, not only about how an answer is worded. Ask the same recommendation question in two languages and a different shortlist of businesses can come back. The language of the question changes who is in the answer, not just the phrasing around it.

Why this matters to a small business

If your customers ask an assistant for a recommendation rather than working through search results and opening pages one by one, the answer is the shortlist they act on.

AI assistant brand recommendations are not one thing you can check once and settle. They come back per language. A customer who asks in Arabic does not see the same list as a customer who asks in English, and neither of them necessarily sees the list you would expect from your own website. A check in one language can look fine while a customer asking in another is shown nothing useful about you.

Which languages matter is a question about your own customers. If everyone who buys from you asks in one language, a check in that language is the whole picture for you. If part of your market asks in another language, then checking only in English tells you about part of your market.

Why AI recommendations in different languages differ

An assistant answers from material it can find in the language the question was asked in, so the same business can look well covered in one language and barely present in another.

We cannot see inside these systems, and neither can anyone outside them. What we can say is what the mechanics suggest: the assistant needs something to read. If little is written about your business in that language, there is little for it to draw on, so it falls back on whatever else it finds — a directory entry written by somebody else, a profile nobody has updated, a description that no longer matches what you do. There is no switch to flip that corrects it.

That is how AI assistants choose brands: not from a ranking anyone can look up, but from what is readable in that language on the day somebody asks.

The gap shows up in one of two ways. Either you are left out of the list, or you are in it with the wrong services, the wrong area, or a name that does not match your own website.

What to do if you are missing in one language

If the answers in a language leave you out, the place to start is whether there is anything for an assistant to read in that language.

• Decide which languages your customers actually ask their questions in. That is the list worth testing, whether it is one language or several.
• If the material in that language is thin, the fix is content: pages, written in that language, that describe what you do.
• Keep your business name, your services and the areas you cover the same in every language you publish in. Two versions that read like two different businesses are harder for these systems to treat as one.
• Correct what is wrong wherever it appears: services you no longer offer, an address you have moved from, a phone number that has changed. Assistants repeat what they find.
• Then wait, and ask the same questions again. New content in a language takes time to appear in answers, in our experience.

Multilingual AI search visibility is mostly this: making sure the material exists, and says the same thing, in each language.

Anything beyond reading and asking — adding pages, editing what is published, changing settings on your site — is a job for whoever looks after your website.

What you can check yourself

Checking your visibility in another language is asking questions and writing down what comes back. None of it changes your website, your accounts or your settings, so you can do all of it yourself.

• Ask an assistant a recommendation question in English — the kind a customer would ask about your type of business in your area — and write down the names it gives.
• Ask the same question in the other language your customers use. If you do not write it comfortably, ask a colleague or a customer to type it, then put the two lists side by side.
• Repeat with two or three questions you actually hear from customers, including the service they want and the area they are in. A general question and a specific one can come back differently.
• Read the answers for accuracy, not only for whether you appear: your name, what you do, where you work, and how someone reaches you.
• Open your own website and see whether it has real pages in that language or only a translated menu.
• Run the same questions in a second assistant. Two systems can answer differently, and it is worth seeing whether the gap looks the same in both.
• Write down the date you checked and what came back, so you can compare it later.

A single answer is not a verdict. Assistants word things differently from one attempt to the next, so the pattern across several questions tells you more than any one reply.

What to do next

• Keep the note of what you asked and what came back.
• If you are absent from the answers in one language, or described wrongly in them, ask whoever looks after your website what content exists in that language today and what could be added.
• If you would rather not work it out alone, ask them to look at it with you.

This is the kind of work our engineers do for clients: checking how a business appears in each language its customers use, and fixing what the assistants are reading.