AI visibility in Montréal and Québec, in French and in English

I work from Brussels, a city that answers in two languages. Montréal does the same thing, with a twist: the language of the question changes who the AIs recommend.

On 30 September 2026, I asked seven AIs the same buyer questions about five professions in Montréal, once in French and once in English, with web search on and without it. This page shows what they answered. I repeat the measurement every month.

This page is for marketing agencies and consultants who work with local businesses in Montréal, Québec City, Laval, Gatineau or anywhere in Canada. It is also for a business owner who wants to know one thing: when a buyer asks an AI, do they name me, in both languages?

One target is one city, in one language, for one type of buyer. Montréal in French and Montréal in English are two targets. I measure them the same day, with the same 24 questions.

For each target, 8 questions a buyer actually asks (who should I call, what does it cost, who has proven results), put to 7 AIs: ChatGPT, Claude, Gemini, Grok, Perplexity, Mistral, Muse Spark from Meta. Each AI answers twice: with web search on, and without it, from what it already knows. That is up to 56 answers per mode. I count who gets named, which sources the AIs read, and which facts they get wrong.

Montréal is the example I publish, not the limit of what I measure. The table below comes from the same measurement as the French page.

ProfessionLanguageModeAnswers receivedAudit ID
DentistefrançaisWith web search56cmuntwz2p006gzdmbzy5l6mw8
DentistefrançaisWithout web search56cmuntwtpa004izdmbrrinjym4
DentisteanglaisWith web search56cmuntwna6002kzdmb350h2842
DentisteanglaisWithout web search55cmuntwg6o000mzdmbkozu3nyt
AvocatfrançaisWith web search54cmuntxlbv00e8zdmbazvjolf6
AvocatfrançaisWithout web search56cmuntxg4x00cazdmbezs9mtmn
AvocatanglaisWith web search56cmuntxbf400aczdmbts2zf6da
AvocatanglaisWithout web search54cmuntx5j7008ezdmblyuc674k
Expert-comptablefrançaisWith web search56cmunty55l00m0zdmbsgb0qoyy
Expert-comptablefrançaisWithout web search56cmunty0kw00k2zdmbhyhvzrrw
Expert-comptableanglaisWith web search54cmuntxvwg00i4zdmbtoos5hfe
Expert-comptableanglaisWithout web search54cmuntxq7900g6zdmbgh4s5qys
Agent immobilierfrançaisWith web search56cmuntynuv00tszdmbah3rmyye
Agent immobilierfrançaisWithout web search56cmuntyis500ruzdmb1pkcq5zb
Agent immobilieranglaisWith web search56cmuntye4v00pwzdmbk49vcuvf
Agent immobilieranglaisWithout web search56cmunty9nn00nyzdmb7bfg6y97
Courtier en assurancefrançaisWith web search52cmuntzbln011kzdmbb43gbo9l
Courtier en assurancefrançaisWithout web search56cmuntz6rx00zmzdmbitnzeq2w
Courtier en assuranceanglaisWith web search56cmuntz0l000xozdmbw1xms2im
Courtier en assuranceanglaisWithout web search56cmuntyv8k00vqzdmbd9d82zru

Three things I read in these answers:

  • For "insurance broker", with web search on, Qubit Insurance appears in 29 % of the English answers (16 of 56) and in 0 % of the French ones (0 of 52). Same city, same question.
  • For "dentist", with web search on, Centre Dentaire Oralvie leads the French answers with 21 % (12 of 56) and gets 7 % in English (4 of 56). Without web search, Centre Dentaire Mont-Royal appears in 11 % of the English answers (6 of 55) and never in French (0 of 56).
  • For "chartered accountant", without web search, MNP leads in English with 37 % (20 of 54) and gets 9 % in French (5 of 56). Raymond Chabot Grant Thornton leads in French with 36 % (20 of 56).

A business visible in one language can be absent in the other. In Montréal, that is half of the buyers.

Any Canadian city can be measured, in English or in French: Toronto, Vancouver, Calgary, Ottawa and Gatineau, Québec City, Sherbrooke. A city I have not measured yet takes one run. The first results arrive within 48 hours.

The Brand Snapshot costs $250, once. It covers two targets, for example Montréal in French and Montréal in English, with a re-measurement of the same questions 30 days later. You receive a report with the answers, the errors the AIs make about the business, the sources they read, and the actions to take, each with a ready-to-paste prompt.

Prices in USD. I bill Canadian businesses in USD; Canadian cards are accepted.

After the Snapshot, the monthly follow-up continues through a partner agency or consultant: $59 per month for the first client, $40 for each additional client, no commitment.

Three things are enough: the name of the business, its website, and the two targets (city and language). You can order the Snapshot here: /snapshot. If you are an agency with several clients, write to me first: /contact?topic=partnership.

VectorGap is a Belgian company, founded in May 2026 in Brussels. I serve Canada remotely, in French and in English. Support by e-mail in your language, by phone or video call in French and English, with a personal answer in under 12 hours.

Frequently asked questions

Can you measure Toronto or Vancouver?

Yes. Any Canadian city, in English or in French. Montréal is the example I published because it lives in two languages. A new city takes one measurement run; the first results arrive within 48 hours.

Do you bill in Canadian dollars?

No. Prices are in USD, the same figure as in euros. Canadian cards are accepted and your invoice shows USD.

Do you measure Québec French and Canadian English separately?

Yes. One target is one city in one language. Montréal in French and Montréal in English are two targets, measured the same day with the same 24 questions. The Snapshot covers both.

Which AIs do you ask?

ChatGPT, Claude, Gemini, Grok, Perplexity, Mistral, Muse Spark from Meta. Each one answers with web search on and without it, so you see what the AI reads today and what it already believes.

Do you need access to my client's website or analytics?

No. I only need the business name, its website, the city and the language. The AIs answer; I read their answers and check the facts against the business's own pages.

These are the answers of the AIs on one day, not a ranking of quality. From one measurement to the next, a difference of less than 15 points out of 100 can come from the AIs' own randomness; I write that in every report, so nobody reads a movement that does not exist.