Same city, same question. Change the language and AI names someone else.
In Montréal, I asked 7 AI models with web search for an insurance broker. In English, Qubit Insurance came up in 16 of 56 answers. In French, in none of 52.
Same business, another language
Answers naming the business, out of answers received. With web search. Each count carries its source audit ID.
| City | Business | Answers naming it | Measured on |
|---|---|---|---|
| Montréal | Qubit Insurance | English16/56 French0/52 | 30 September 2026 |
| Brussels | Yago | English2/56 French29/56 Dutch8/53 | 24 September 2026 |
| Luxembourg | Weicker & Co | English6/50 French20/50 German7/53 | 24 September 2026 |
| Paris | Laforêt | English6/54 French18/50 | 24 September 2026 |
| Paris | Paris Property Group | English13/54 French0/50 | 24 September 2026 |
Measured 24 to 30 September 2026. A citation in an answer is a count, not a quality ranking.
What the table shows
- A business can lead in one language and be missing in another, in the same city.
- In Paris it works both ways. Laforêt is named in 18 of 50 French answers and 6 of 54 English ones. Paris Property Group is named in 13 of 54 English answers and in none of the French ones.
- In Luxembourg each language puts a different broker first: Weicker & Co in French, OCA in English, IBG in German.
If your client sells in more than one language
One Snapshot covers two targets: two languages, two personas or two places. For a business that sells in two languages, I measure both, in the same 24 hours. 30 days later, I run the same questions again: that is the 30-day proof loop.
$250USD · one-time payment
Delivered within 48 hours of your completed brief.
10-minute video walkthrough included. No subscription. VectorGap-branded report. Taxes may apply.
One Snapshot covers two targets: two languages, two personas or two places.
The second target changes only one element: the language, the persona or the place.
The 30-day proof loop: a second run of the same questions on the same models, 30 days after the report, with what moved. The 30-day proof loop covers both targets.
How I measured
- The same standard buyer questions, asked in each language.
- The panel at the time: ChatGPT, Claude, Gemini, Perplexity, Grok, Mistral, DeepSeek. With web search.
- The figure is the number of answers naming the business, out of answers received.
- Measured 24 to 30 September 2026. How we measure.
- The current panel uses Muse Spark in place of DeepSeek.
Use this data
Download the CSV, with counts and audit IDs
License: CC BY 4.0.
VectorGap, Multilingual AI visibility measurements, September 2026, vectorgap.ai/multilingual-ai-visibility
