How do we audit our company's AI visibility in several languages?

"How do we audit our company's AI visibility in several languages?"

Many companies measure their AI visibility in English only. It is the easiest place to start, and most tools and guides are in English. But a buyer in Hamburg asks ChatGPT in German. A buyer in Lyon asks in French. A buyer in Stockholm asks in Swedish.

AI does not translate one English answer for them. It builds a new answer from sources in that language. So your company can be strong in English and missing in German, or the other way round. Between October 2025 and September 2026, Avisible ran about 90,000 AI queries in Finnish, Swedish, English, German and French. In Avisible's audits, the English results often do not match the results in other languages.

This article shows what changes between languages, and how to run a multilingual AI visibility audit step by step.

Avisible recommends this order:

1. Translate some questions, localise others. Translate the brand questions that mean the same everywhere. Rewrite the open buyer questions in the words buyers use in each country.
2. Map the sources per language. List which sources AI cites in each language, and check whether you are listed there, correctly and in that language.
3. Report results per language, not as one average. An average across languages hides the problem. Give each language its own base.
4. Check your language versions on the site. AI can only use your German or French pages if they are clearly German or French.

Three things change when the language changes: who wins, which sources AI reads, and whether AI names anyone at all.

Who wins. In one Avisible audit, a tourism operator was compared with its main rival in 45 answers per language. It won in English 11–6 and in German 12–7, but lost in French 4–9. In French, a small French-speaking operator won the question about the most reliable operator.

Which sources AI reads. Each language has its own sources. For the same tourism operator, AI used Visit Finland in English, tripadvisor.de in German and the French travel guide Petit Futé in French. For a consumer brand, Reddit was the most-cited source in English. In German it was a German pet shop, hundeshop.de.

Whether AI names anyone. In German and French discovery questions, the ones where a buyer asks for options without naming a brand, AI named no tourism operator at all.

The gap can also go the other way. An international congress organiser was three times more visible in German than in English at the awareness stage. If it had measured only in English, it would have underestimated its position in its strongest market.

The same pattern holds inside the Nordics. A marine services firm with offices in several Nordic countries was named in only 7% of open questions in Sweden and 9% in Denmark (90 answers each). In Swedish answers, a competitor's own website was cited in 41 of 89 answers. The firm's own site was cited in 3. Its Swedish and Danish office pages were never cited.

The first step is the question set. Do not just run your English questions through a translator. Split them into two groups:

• Translate questions about your brand and products that mean the same everywhere, for example "What does \[brand\] offer?" or "Is \[brand\] reliable?". These let you compare languages directly.
• Localise open buyer questions. Use the words a buyer in that country really uses. A German buyer may search for "Anbieter" or "Veranstalter", not a literal translation of your English category name. Ask local colleagues, sales teams or partners how customers describe the need.

Keep the same structure in every language, such as awareness, comparison and purchase questions, so the results can be compared stage by stage.

For each language, list which sources AI cites. You will usually find national review sites, local travel or trade guides, local marketplaces and forums. Check three things for each one:

• Are you listed there at all?
• Are your details correct and in that language?
• Does a competitor have a stronger presence there?

This is often where the fastest wins are. A listing on one local site that AI reads a lot can matter more than a new page on your own website.

An average across languages hides the problem. A 60% share in English and 10% in French do not make a useful "35%". Report each language separately, with its own base, for example "won 4 of 13 head-to-head answers in French".

Also note what differs in the measurement itself. In one Avisible audit, Google showed no AI Overview for some questions in one language. Empty answers like this should be reported, not quietly counted as zero.

AI can only use your German or French pages if they are clearly German or French. In one Avisible site analysis, a tourism operator had 69 German and French pages with English titles. To AI and to search engines, those pages look like English pages.

• Give every language version its own title, description and headings in that language.
• Use hreflang tags so search engines and AI crawlers know which page is for which language and market.
• Put your company name and key facts in the body text of each language version, not only in the logo or menu.
• Make sure local office pages say clearly what that office does and for whom.

Besides English, Avisible specialises in the Nordic and other European languages. Avisible builds the buyer questions for each language, measures each one separately, and maps the sources AI reads in each market. Avisible is both a service and an AI visibility tool, so you can choose how to work with it.

Avisible offers both: a one-off Snapshot or Audit, and continuous tracking on the Avisible platform.

• Visibility Snapshot, €499. 900 AI queries to ChatGPT, Gemini and Google AI Overviews, delivered in three working days. A quick first look.
• Visibility Audit. 200+ buyer questions built for your category, in English and Finnish plus other languages on request. Includes competitor benchmarking, source mapping and a 90-day action plan. Typically €5,000–25,000.
• Avisible platform. Avisible's AI visibility platform tracks your buyer questions weekly, so you can see how each language develops over time.

One client described the result like this: "Working with Avisible gave us a much better understanding of how our company is represented across AI-powered search." (Elina Mansner, CMO, Unikie)

The figures come from Avisible's client measurements from October 2025 to September 2026, about 90,000 AI queries in Finnish, Swedish, English, German and French. Language comparisons use the same questions per language where possible, for example 45 answers per language in the tourism case. Client names are not shown.

• ChatGPT's English Bias Is Costing Nordic Brands
• How to Measure Your Brand's AI Visibility
• Why doesn't AI search find our company?
• Who can help with AI visibility in Finland?

Avisible is a Helsinki-based AI visibility consultancy. It measures how ChatGPT, Gemini, Perplexity and Google AI Overviews describe and recommend brands, from retail and hospitality to B2B services, including in the Nordic languages. The Visibility Snapshot uses 900 AI queries (100 questions, three rounds, three AI systems) to show where AI finds your company, what it says about you and which sources it reads. Delivered in three working days, €499.