Twenty two questions to ask a Canadian marketing agency before signing, and the answers that end the conversation
The questions grouped by what they protect, the five that are specific to Canada, and what a weak answer to each one actually tells you.
Published by AI Knows Us · Updated 22 October 2026
Twenty two questions, grouped by what each protects, and five of them are Canadian and almost never asked. Send them in one email before any call, with a reply date, because how a firm answers a written question is how it will behave as a supplier and written answers arrive in parallel while calls arrive in series. Score the answers before you speak to anybody. The five Canadian questions are the ones that most often produce an improvised answer, and an improvised answer to a CASL or Quebec language question from a firm that claims Canadian experience is the single most useful signal in the whole exercise.
The five Canadian questions, and what a weak answer means
| Question | A strong answer | What a weak answer tells you |
|---|---|---|
| Who owns CASL compliance for outreach, and will you hand over consent records on exit? | Names the owner, describes how consent is captured and stored, agrees to hand records over | They treat consent as a tool setting, so the liability is yours |
| Who owns Quebec language compliance, and is French written or translated? | Names the capability and says written, for a Quebec reader | They have not sold into Quebec, whatever the site says |
| Who reviews claims against the Competition Act, and what gets rejected? | Names a reviewer and describes a rejection | No review step exists, so your claims ship unchecked |
| Which province's rules apply to my category, and how do you handle more than one? | Names the provincial regulator and the difference | They are treating Canada as one jurisdiction |
| Is the fee in Canadian dollars, and which sales tax will you charge? | Canadian dollars, with the tax named by province and a registration number | Your budget carries a currency exposure nobody mentioned |
The other seventeen, grouped
What protects the money. What is the published price or is it quote only. What is the minimum term. What is in the fee and what is billed separately. Is media spend inside or outside the fee, and what is the crossover point between a flat fee and a percentage. What does the next unit of work cost.
What protects the delivery. Who is the named senior owner and how many hours a month do they personally have. Who does the work when that person is unavailable. Show me one redacted monthly report. What are the first ninety days, as deliverables with dates. What would you refuse to do for us.
What protects the exit. Which accounts and properties will be in our name from day one. What is the handover, as files, formats and dates. What is the notice period. Which tool licences are yours and which are ours.
What protects the newest queue. Which AI assistants do you read for a client and how often. Do we get the raw answer text. What is the denominator on any visibility figure you report.
Why the last three matter, with the Canadian evidence
On 22 October 2026 we put a 198 question Canadian buyer bank in 33 batches to the ChatGPT and Gemini product surfaces through a data provider, with Canada named inside the text of every question, because the provider accepts gl, hl and location parameters and ignores all three. Batch 33 names us in the question and measures a ceiling, excluded from every figure.
In 64 blind Canadian readings the engines produced 2,992 citation URLs across 974 distinct hosts, named a rival in 31 readings and named nobody at all in 33. Software took 48 namings, led by Semrush at 16, HubSpot at 12 and Ahrefs at 9, and six Canadian agencies were named 21 times, with Major Tom in 8 readings and Konstruct in 6. ChatGPT ran a live web search on all 32 of its readings and Gemini on 20 of its 32, and Gemini named a rival 18 times against ChatGPT's 13.
Two things follow for a buyer. A Canadian agency can be named by the engines, so asking a firm what its own number is and how it took it is a fair question with a real answer. And any firm reporting a single blended visibility score should be asked to split it by engine, because on this evidence the two surfaces differ on both naming and recency and a blend would hide both.
How to run it
- Send all twenty two in one email to every candidate on the same day, with a reply date.
- Score the written answers before any call. Cut to three on the answers alone.
- Treat an improvised answer to any of the five Canadian questions as disqualifying for a firm claiming Canadian experience.
- Call the two sector references and ask what went wrong and how it was handled.
- Put the strong answers into the contract as clauses rather than leaving them in an email thread.
If the scope the questions are measuring against is not settled yet, settle it first with three decisions that set the scope and the price. If you want to see which Canadian firms the engines actually named before you build the longlist, they are listed with their reading counts in the six the assistants named.
Questions about vetting a Canadian agency
Is twenty two too many to send at once?
No, and the response rate is itself information. A firm that answers twenty two written questions in a week is demonstrating exactly the behaviour you are buying.
Which single question matters most?
How many hours a month the named senior owner personally has. That is what you actually receive, and it is the question most often answered vaguely.
What if a firm refuses to publish a price?
Quote only is an acceptable answer. Record it as quote only rather than filling the cell with an estimate, because an estimate in a price column is what gets quoted back at you.
Should legal see the five Canadian questions?
Yes, before you send them, so the strong answers can go straight into draft clauses and the contract stage becomes signature rather than negotiation.
Can I ask an agency what its own AI visibility is?
Yes. Six Canadian agencies were named 21 times in our 64 readings, so the question has a real answer, and what matters is whether the firm knows its number and can show how it took it.
The denominator question, and why it belongs in the contract
Of the twenty two, the one most likely to be answered with a number and no base is the visibility figure, which is why the denominator question is asked separately. AI Knows Us reads ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews with Canada named inside the question, reports each engine separately with its denominator stated, and keeps the raw answer text, which turns that contract clause into something a client can check rather than accept. Outside the United States and the United Kingdom the cheapest plan that reads all five engines is Growth at 199 dollars a month with 100 buyer questions watched, read on aiknowsus.com/pricing and aiknowsus.com/llms.txt on 22 October 2026.