AI visibility for Canadian law firms: FLSC Rule 4.2-1, provincial rules, CASL and Law 25 in one guide

What rules govern AI visibility work for a Canadian law firm?

Four layers apply. The provincial law society code governs every public statement the firm makes about itself, including the pages AI engines quote. Law society guidance on generative AI governs how that copy is drafted and checked. CASL, the federal anti-spam law, governs commercial email and text sent to a lead. Privacy law, with Quebec's Law 25 the strictest, governs the analytics that measure AI referrals.

The American version of this question is mostly a single-layer problem: ABA Model Rule 7.1 and the state filing regimes. The Canadian version is wider because the follow-up and measurement layers carry federal and provincial statutes with penalties that dwarf a law society reprimand. A firm that wins an AI citation and then emails the lead without valid consent has turned a marketing gain into a regulatory exposure.

The upside is consistency. Every common-law province builds its marketing chapter on the Federation of Law Societies of Canada Model Code, so the core test is the same from Halifax to Victoria. Quebec runs its own code. The differences that matter for AI visibility fit in one table, which is section three.

What does FLSC Model Code Rule 4.2-1 require?

FLSC Model Code Rule 4.2-1 permits a lawyer to market professional services if the marketing is demonstrably true, accurate and verifiable, is not misleading, confusing or deceptive, and is in the best interests of the public. According to the Canadian Bar Association's advertising toolkit, the commentary lists specific contraventions: suggesting qualitative superiority, raising expectations unjustifiably, and stating amounts recovered without a past-results qualifier.

That last item is the one AI visibility work runs into most. The commentary treats a statement of money recovered or degree of success as a contravention unless it is accompanied by a statement that past results are not indicative of future results. On a web page, that qualifier usually sits in a footer or a separate disclaimer block. An AI engine does not quote the footer. It lifts one sentence.

So the drafting standard for a Canadian page is the quoted-alone test: every sentence that states a result must still comply if an engine lifts it with nothing around it. In practice that means writing the qualifier into the same sentence as the result, or not stating the result in a citable sentence at all. Firm AEO treats a results sentence that fails the quoted-alone test as non-compliant copy, whatever the page footer says.

The same commentary bars suggesting the lawyer is aggressive and using testimonials or endorsements that contain emotional appeals. Both show up in practice-page copy written for conversion. Both get lifted by engines answering situational questions, which is exactly where a frightened reader is most open to an unjustified expectation.

How do the provincial rules differ?

The common-law provinces share the Model Code test and differ mainly on three points: how strictly the regulator reads superlatives and rankings, how the word specialist is controlled, and how much generative AI guidance has been published. Quebec differs on all three and adds a full ban on testimonials. The table below covers the five largest legal markets; Atlantic and prairie law societies follow the Model Code chapter 4 with local variations.

Ontario is the province where AI answers and marketing rules collide most directly. The Canadian Bar Association toolkit reports that Ontario prohibits comparative language such as better, greatest or fastest, and treats claims to be the best as a contravention unless an objective, valid process stands behind them. British Columbia's Rule 4.2-5 bars marketing that is false, inaccurate, unverifiable or reasonably capable of misleading, and its commentary bars creating an unjustified expectation about results.

Can a Canadian law firm describe itself as the best when AI engines ask for the best?

No. AI engines answer prompts phrased as "who is the best criminal defence lawyer in Toronto", but the Law Society of Ontario treats "best" as a contravention when a firm says it about itself. The compliant route is third-party corroboration: bona fide rankings, law society directory listings and press coverage that an engine reads and combines. The firm states verifiable facts. The engine does the ranking.

This is not a disadvantage peculiar to Canada. American firms under ABA Model Rule 7.1 face a softer version of the same limit, and the engines rarely quote a firm's own superlatives anyway, because self-description is the least corroborated claim on the web. What the Canadian rule changes is the copy audit. A page that says "Toronto's top family lawyers" is both non-compliant and uncitable, so removing it costs nothing in visibility.

What a firm can publish is the verifiable fact underneath the superlative. Years of call to the bar, the courts the lawyer appears in, a Lexpert or Best Lawyers in Canada listing named with its year, the specific limitation period that applies to the reader's matter in that province. Each of those is a sentence an engine can lift and a regulator can verify. The Canadian corroboration sources themselves are mapped in Firm AEO's earlier post on Canadian citations.

What do the law societies say about using generative AI to write marketing copy?

Every major Canadian law society that has published generative AI guidance treats the lawyer as responsible for AI-drafted output, and at least two name marketing directly. The Law Society of Alberta's Generative AI Playbook, updated February 2026, lists website content, blog posts and marketing emails as uses. The Law Society of Ontario's April 2024 white paper lists marketing as a licensee use case.

None of that guidance requires a label on AI-assisted marketing copy, and Firm AEO has not found a Canadian law society that does. What the guidance requires is verification: competence, confidentiality and supervision apply to a practice page the same way they apply to a factum. A practice page drafted with an AI tool that states the wrong limitation period, or invents a result, is the firm's statement under Rule 4.2-1.

The enforcement record shows how seriously verification failures are treated, even though the cases so far involve court filings rather than marketing. In Zhang v. Chen, 2024 BCSC 285, the British Columbia Supreme Court ordered costs personally against counsel who filed ChatGPT-fabricated cases. In Law Society of Ontario v. Lee, 2026 ONLSTH 136, the Law Society Tribunal imposed a six-month suspension and $10,000 in costs after a factum cited non-existent cases and the lawyer denied using AI. Law Times reported a separate decision, 2026 ONLSTH 112 of June 17, 2026, ordering $31,150 in costs against a lawyer who filed AI-fabricated citations.

• •Check every statute, limitation period and court name on an AI-drafted page against the primary source before it publishes.
• •Strip any result, superlative or expectation the drafting tool added, then run the quoted-alone test on each remaining sentence.
• •Keep client facts out of the prompts used to draft marketing copy; confidentiality applies to the drafting tool as well as the page.
• •Record who reviewed the page and when, so the firm can show supervision if a regulator asks.

How does CASL apply to leads that arrive from ChatGPT?

CASL applies the moment the firm sends a commercial electronic message to the lead. A visitor who arrives from ChatGPT and submits an inquiry form gives implied consent for 6 months from the inquiry, according to Innovation, Science and Economic Development Canada. A retained client creates an existing business relationship that supports implied consent for up to 2 years. Express consent has no time limit until withdrawn.

The trap is the nurture sequence. American intake platforms often ship a default sequence that emails a lead for months after the first contact. In Canada, a six-month sequence that keeps running past the inquiry window on implied consent alone is sending without consent. Every commercial message also has to identify the sender and carry a working unsubscribe mechanism, and the burden of proving consent sits with the sender.

The stakes are high relative to the size of most firms. Administrative monetary penalties under CASL reach up to $1 million per violation for individuals and up to $10 million for organizations, according to the Canadian legal information summary at law-faqs.org. For an AI-referred lead, the practical fix is an unticked express-consent checkbox on the intake form, worded separately from the inquiry itself, with the consent date stored in the CRM.

What does Quebec's Law 25 change about measuring AI referrals?

Law 25 requires prior consent for non-essential cookies and tracking, and the requirement follows the visitor's location rather than the firm's. A Toronto firm with Montreal visitors is in scope for those visitors. Any analytics tag that sets non-essential cookies before consent is a compliance problem, and any analytics tag that waits for consent undercounts. AI referral measurement for Quebec traffic will read low.

The penalties set the priority. According to Osler, the enforcement scheme in force since September 22, 2023 allows administrative monetary penalties of up to C$10 million or 2% of worldwide turnover, whichever is greater, and penal fines of up to C$25 million or 4%. No AI referral dashboard is worth that exposure, so the consent banner wins and the measurement method adapts.

The adaptation is to measure upstream and at intake rather than in analytics. Firm AEO measures Quebec AI visibility by running the client prompts directly across the engines in French and English, recording which firms are named and which sources are cited. Attribution moves to a "how did you find us" field on the intake form, with ChatGPT, Google AI and Perplexity as named options. Neither method depends on a cookie.

Which rule governs each AI visibility task?

Most AI visibility tasks touch exactly one Canadian rule, which makes the compliance review short once the map exists. The table below pairs each task Firm AEO performs for a Canadian firm with the rule that governs it and the practical constraint. It is a planning document for the firm's ethics counsel and privacy lead, not a substitute for their review.

In what order should a Canadian firm work through these rules?

Fix the follow-up and measurement layers first, because they carry the largest penalties and the fastest fixes, then audit the page copy, then build corroboration. The sequence below is the typical order Firm AEO uses with Canadian firms; timings are illustrative and depend on the size of the site and how quickly the firm's ethics counsel reviews.

• •1. Add a separate express-consent checkbox to every intake form and stop any nurture sequence running on expired implied consent (typically one week).
• •2. Confirm the consent banner blocks non-essential tracking for Quebec visitors and add a "how did you find us" intake field with AI engines as options (typically one week).
• •3. Audit every practice page for superlatives, specialist language, aggressive framing and results sentences that fail the quoted-alone test (typically two to four weeks).
• •4. Replace American terminology with the province's: limitation period, criminal defence, the correct court names (often done in the same pass as step 3).
• •5. Set a written review step for AI-drafted copy, naming the reviewing lawyer and recording the date.
• •6. Build third-party corroboration through law society directory accuracy, Canadian rankings and Canadian legal press, then measure monthly across the engines.