Can a Canadian law firm get cited in ChatGPT or Perplexity?

Why does none of the published citation data describe Canada?

Because every study of AI legal citation published so far sampled American queries in American cities. Citorian ran 359 answers across ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews in June 2026, in New York, Los Angeles, Chicago, Houston and Phoenix. 5WPR and Haute Lawyer tested American practice areas against American directories in April 2026. A Toronto, Calgary or Halifax firm reading either number is reading a foreign map.

The consumer-side figure Canadian agencies quote most has the same problem. The iLawyerMarketing survey published on August 17, 2026 that put ChatGPT use for lawyer research at 41.9% sampled 1,110 US adults. Firm AEO has found no Canadian equivalent. The closest citable source is Clio’s 2025 Legal Trends Report, from a Canadian company with a North American sample, which reported that more than half of consumers have used or would consider using AI to answer a legal question, and that 28% of those who used AI were directed to contact a lawyer.

What replaces the US directory layer in Canada?

No single Canadian source carries what Super Lawyers carries in the United States, where it took 33% of all citations in the Citorian sample. The replacement is a set: the provincial law society directories, Lexpert and the Canadian Legal Lexpert Directory, Best Lawyers in Canada, CanLII, the Canadian Bar Association, and the Canadian legal press. Firm AEO treats that set as a hypothesis to test per city, not a ranked roster.

The split inside the American roster is the part most Canadian firms get wrong. Of the seven directories that 5WPR and Haute Lawyer identified as owning the legal citation layer, the credentialing titles publish real Canadian editions, so a Bay Street or Howe Street commercial practice is already represented in the sources the engines read. The consumer directories are the ones that stop at the border. That inverts the American picture: in Canada, the family, criminal defence and personal injury practices have the larger corroboration gap, not the commercial ones.

Which engines are most likely to name a Canadian firm?

The Google-fed engines. Martindale-Avvo reported in April 2026 that ChatGPT matches Google’s top ten results for legal queries less than 25% of the time, while Perplexity and Claude sit near 75% and Gemini near 50%. A Canadian firm that ranks in Google for a practice and a city therefore carries into Gemini, Claude, Perplexity and Google AI Overviews with comparatively little new work.

ChatGPT is the hard one, and it is hard for a structural reason rather than a Canadian one. Martindale-Avvo named Super Lawyers, Avvo, Martindale-Hubbell and FindLaw as the four sources ChatGPT cites most for legal queries. Three of those four have little Canadian consumer depth. The engine with the largest audience is reading the part of the roster that stops at the border, which is why Canadian consumer firms are absent from ChatGPT more often than their Google position predicts.

Absence is the default everywhere, not a Canadian penalty. The SOCi 2026 Local Visibility Index, published February 17, 2026 across 2,751 brands and roughly 350,000 locations, found ChatGPT recommending 1.2% of locations, Perplexity 7.4% and Gemini 11%, against 35.9% in Google’s local three-pack. A Canadian firm starting from zero is starting where almost every business starts.

Why do Canadian marketing rules make corroboration the only route?

Because AI engines answer questions phrased as “who is the best employment lawyer in Toronto”, and a Canadian firm cannot answer that question about itself. FLSC Model Code Rule 4.2-1 requires marketing to be demonstrably true, accurate and verifiable, and not misleading, confusing or deceptive. The Law Society of Ontario commentary on the same rule treats “best”, “super” and “#1” as contraventions. What a firm cannot claim, a third party has to establish.

Quebec is stricter again. Article 145 of the Code de déontologie des avocats bans testimonials and endorsements outright, so a Quebec firm republishing Google reviews on its own site has a problem even though those same reviews feed the engines that recommend it. In British Columbia, chapter 4 of the Law Society of British Columbia Code bars marketing that raises an unjustified expectation about results and states that websites and social media are not a softer medium. Firm AEO explains these as marketing constraints and does not give legal advice. The firm’s ethics counsel decides what the firm publishes.

Does Canadian legal terminology change what gets retrieved?

It changes which page the engine finds. A Canadian client asks how long they have to sue, and the answer on the page has to say “limitation period”, because “statute of limitations” is the American phrasing and the two strings retrieve different documents. The same applies to “criminal defence” against “criminal defense”, to the Insurance Act and the Statutory Accident Benefits Schedule in Ontario motor vehicle work, and to the correct provincial court names.

This is the cheapest Canadian correction available and the one most often skipped, because a page written on an American template reads as competent and still misses the retrieval. Firm AEO checks practice pages for American phrasing first, since a page that does not match the question is not a candidate for citation however strong the firm’s directory profile is.

How should a Canadian firm test the map for its own city?

By running the Citorian method against Canadian cities and Canadian phrasing. The exercise takes an afternoon and produces the only Canadian citation data that exists for a given firm, which is its own. The output that matters is not which firms were named. It is the list of domains the engines cited, because that list is the corroboration map for that city and it replaces the American roster in the table above.

Canadian measurement carries one extra problem worth handling at the same time. Quebec’s Law 25 requires prior consent for non-essential tracking, and consent applies by visitor location rather than business location, so Quebec traffic is under-counted before any AI referral question arises. Most AI referrals arrive without a usable referrer in any province. Firm AEO moves attribution into a “how did you find us” field on the intake form rather than trying to recover it from analytics.

• •Write ten prompts a client would actually type, in Canadian terms: “limitation period” rather than “statute of limitations”, “criminal defence” rather than “criminal defense”, the province named, the city named.
• •Run each prompt three times in ChatGPT, Perplexity, Gemini, Claude and Google AI Mode, logged out with history disabled, because a signed-in session returns a personalised answer that cannot be compared month to month.
• •Record two columns for every answer: which firms were named, and which domains were cited.
• •Count the domain column. Whatever appears three or more times is the corroboration set the engines are actually using for that city.
• •Repeat monthly. Weekly runs measure sampling noise rather than change, and the engines do not re-crawl fast enough to justify them.