The Canadian data on assistant use before a purchase, and what each kind of source actually measured

Canadian figures on assistant use before a purchase exist and they are weaker than their headlines, because they come from five different kinds of source and each measures something different. National statistical collection, telecommunications and internet usage reporting, consumer surveys, platform disclosures, and vendor commissioned research. Those five disagree, and the biggest single reason is definitional: whether an AI answer inside a search results page counts as using an assistant. So the useful posture for a Canadian business is to read any figure with four questions attached, and to take one measurement of its own, because nobody else can measure what the engines say when somebody asks about your category in Canada.

The five kinds of Canadian source, and what each measures

Source type	What it measures	What it cannot tell you
National statistical collection	Technology adoption at a population level, on its own definitions, with a known method	Commercial intent, or what was asked of an assistant
Internet and telecommunications reporting	Access, usage and infrastructure, which bounds the population	Anything about supplier selection behaviour
Consumer surveys	What Canadians say they do	What they did, since routine use is under reported and novel use over reported
Platform disclosures	Usage a platform chooses to publish, on its own definition	Comparability with anything else
Vendor commissioned research	Something real, with a question set designed by an interested party	How the result looks under a neutral question set

The first two are the most reliable in Canada and the least directly useful, because they describe adoption rather than buying behaviour. The last three are the most quoted and the most definition dependent. That combination is why no single Canadian percentage answers this question.

The four questions to attach to any Canadian figure

• What was the exact question wording? Shopping research, supplier selection and general use are three different populations, and the headline usually collapses them.
• Who was sampled, and how recruited? Consumers and business buyers behave differently, and Canada's two language populations are often sampled unevenly.
• Did an AI answer inside a search results page count? This single choice moves the figure substantially and explains most of the disagreement between sources.
• What were the field dates? Not the publication date. This behaviour is moving fast enough that two quarters is a different market.

A Canadian figure with all four attached is worth putting in a board paper, labelled as what it measured. One with none of them is the figure that gets challenged in the meeting, correctly.

The measurement we took, and its denominator

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. Canada was named inside the text of every question, because the provider accepts gl, hl and location parameters and ignores all three, so the prompt is the only market lock that holds. Batch 33 names us in the question, which makes it a ceiling measurement, excluded from every figure.

• 64 blind Canadian readings, 32 per engine, producing 2,992 citation URLs across 974 distinct hosts.
• Named us in 0 readings. Cited our domain in 0 readings.
• Named a rival in 26 readings, 69 namings. Semrush 16, HubSpot 12, Ahrefs 9, Profound 2, Peec 2, Otterly 2, Brandwatch 1, SE Ranking 1, BrightLocal 1, Moz 1, Rankscale 1.
• Named six Canadian agencies unprompted, 21 namings. Major Tom 8, Konstruct 6, Jelly Digital 2, Seota 2, Pilot Digital 2, Edkent 1.
• Named nobody at all in 38 readings, the largest single outcome.
• ChatGPT ran a live web search on 32 of 32 readings. Gemini on 20 of 32. Gemini named a rival 18 times against ChatGPT's 13.

What this measures is what the engines answer, with a denominator and a date. What it does not measure is how many Canadians ask, and we publish no share figure because we took none. Those are two instruments and a page that mixed them would be doing exactly what the five source types above do.

Why the content of the answer matters more than the share

Suppose the true share of Canadian buyers consulting an assistant before contacting a supplier is 15 per cent. Suppose it is 50. In both cases, if the answer those buyers receive names nobody in your category, the action is identical: publish the specific, dated, checkable answer and get cited on the Canadian hosts the engines read. The share changes the urgency and not the action, and the action is knowable from a denominator you control.

What the engines cited, which is the actionable half

Four Canadian federal properties were heavily represented in the citation list: canada.ca in 10 readings, competition-bureau.canada.ca in 9, ised-isde.canada.ca in 9 and www150.statcan.gc.ca in 8. Four Canadian commercial or city domains accounted for 40 readings between them: torontoseo.com 13, canadacreate.com 11, wisemedia.io 9 and ottawaseo.com 7. Google's documentation properties led overall, developers.google.com in 17 and support.google.com in 14. One entry, ccn.web.tr, is the data provider's own fetch infrastructure rather than a publisher and is excluded from every source analysis.

That list is a usable brief in a way a share statistic is not. Canadian official data is free, authoritative and cited, which makes a page built on it with a named table and a date one of the cheapest quotable assets available to a Canadian business.

How to take your own reading

• Write twelve buying questions in the words your Canadian customers use, with Canada or the province named in each, and a French set if Quebec applies.
• Put each to both engines and save the answer text, the named suppliers and the source list, with the date, engine and language.
• Score three ways. Named you, named a rival, named nobody.
• Count the hosts and separate infrastructure entries from publishers.
• Repeat monthly with the same questions and the same denominator, reporting the engines separately.

For a larger single engine source mix before you build your own, the counted version is in the source mix behind 306 readings. If the comparison you actually want is between the two queues rather than a share, it is in what each is better at and what the data shows.

Questions about Canadian assistant usage data

Is there a reliable Canadian figure?

There are several, from five kinds of source, disagreeing because they measure different things. Read any of them with the four questions attached rather than as a single fact.

Why will you not publish a Canadian share figure?

Because we did not measure one. We measured what the engines answered, 64 blind Canadian readings on 22 October 2026, and mixing a measured figure with an estimated one would weaken both.

Which Canadian source type is most trustworthy?

Whichever publishes its question wording, its sample, its definition and its field dates. The disclosure matters more than the type.

Does AI Overviews count as assistant use?

It is a definitional choice rather than a fact, and it is the choice that most moves the published numbers. Decide for your own reporting, write it down and keep it constant.

How small can my own Canadian reading be?

Twelve questions is enough to decide what to publish. It is a census of the questions you chose rather than a sample of a population, so it supports your decisions and not a market claim.

A Canadian denominator you own instead of a headline you borrow

Every published share figure answers a question somebody else framed; a reading answers yours. 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, records every cited host, and keeps the raw answer text so a figure can be audited rather than quoted. 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 and 1,600 credits, read on aiknowsus.com/pricing and aiknowsus.com/llms.txt on 22 October 2026.