Legal Marketing

AI Content and the Law Society: What Canadian Firms Can (and Can't) Publish

15+ years in local marketing; Google Ads certified; Shopify Partner.

You've been pitched. You've read the AI hype. You've probably already had an associate or a paralegal paste something into ChatGPT and ask if they can use it on the firm website.

And you're stuck on one question: does the Law Society actually care? The short answer on ai content law society rules is yes, they care, but not in the way most people think. They don't care that a robot wrote it. They care whether it's accurate, verifiable, not misleading, and whether you've taken responsibility for it before it goes live.

That's the whole frame. Once you get that, the rest of this article is just the specifics by province, plus a workflow you can actually run inside your firm without getting a nastygram from the bencher.

I'm going to cover Ontario, BC, Alberta, and Quebec, because that's where most of my Canadian law firm clients practice. If you're in another province, check your own rules, but most of the country is modelled on the Federation of Law Societies of Canada Model Code, so the principles carry.

The One Rule That Matters Across Every Province

Every Canadian Law Society marketing rule traces back to the same core test. Ontario Rule 4.2-1 says marketing must be "demonstrably true, accurate and verifiable" and "neither misleading, confusing, or deceptive." BC's Code Rule 4.1-1 says the same thing in slightly different words. Alberta's Rule 7.2 says the same thing. Quebec's Code (Rule 147) says the same thing.

Nothing in any of those rules mentions AI. Nothing mentions ChatGPT. Nothing mentions Claude or Gemini or whatever tool your associate is using at 11pm.

Here's the thing. The rule doesn't care how the content got made. It cares whether what you publish is true, accurate, verifiable, and not misleading. If a human wrote a misleading blog post, that's a breach. If AI wrote an accurate blog post and you verified every factual claim before publishing, that's fine.

So the real question isn't "can I use AI?" The real question is: what's your verification workflow, and who signs off before it goes live? That's the piece most firms get wrong.

For a broader look at how this fits into your overall compliance picture, see our complete guide to SEO marketing for lawyers in Canada. This article zooms in on the AI-generated content question specifically.

What Actually Gets Firms in Trouble

In my experience working with law firms across Canada, the AI content problems that blow up at the bar complaint stage usually fall into five buckets. None of them are "we used AI." All of them are verification failures that happened to involve AI.

Fabricated testimonials. Ontario Rule 4.2-1 commentary, read alongside the profession's no-emotional-appeals rules, effectively bans most testimonial content on lawyer websites. If you generate a "client review" with ChatGPT, you've created a fake client who never existed, saying something that was never said. That's not an AI problem. That's a fraud problem. LSO does not care whether Claude wrote it or your cousin wrote it. The firm I heard about that got flagged for an AI testimonial paid roughly $15,000 to sort it out. Don't do this.

Guaranteed outcomes and success rates. Under the Competition Act plus provincial marketing rules, you can't claim "98% success rate" or "we've won every case" unless you can produce verifiable records. AI tools love to generate this kind of puffery because it sounds good. If you let it ship, you're on the hook.

Unauthorized practice and misleading advice. Generic AI-written content that says things like "in Ontario you have two years to file" without the proper qualifications can cross into legal advice territory, and if the generic content is wrong for the reader's actual situation, that's misleading. This is especially risky on FAQ-style pages.

Quebec language parity. If you advertise to Quebec residents, the Charter of the French Language plus Barreau expectations mean you need French content parity. If you run your AI content in English and then auto-translate into French without a francophone lawyer reviewing it, you're shipping potentially inaccurate legal content in a language nobody at your firm reads carefully. That's a problem with or without AI.

Practice specialty claims. Most provinces restrict "specialist" claims unless you hold a certification (Quebec Rule 147 is specific on this). AI content loves to write "our specialists in personal injury law"... which, if you're not certified, is a breach.

None of those are AI problems. They're verification problems. Which means they're solvable.

Province by Province: What Changes

The core principle is the same everywhere, but the edges differ.

Ontario (LSO). Rule 4.2-1 is the anchor. Testimonials are the big one here. Most practitioners read the commentary conservatively and just don't publish client testimonials, full stop. If you're using AI to draft content, the testimonial ban means you should have a hard rule that AI never generates a quote attributed to anyone, real or fake. For a deeper breakdown on the provincial marketing rules generally, see our SEO laws breakdown.

British Columbia (LSBC). BC Code Rule 4.1-1 mirrors Ontario's test. BC is particularly tight on disclaimer language in paid advertising. If you're using AI to draft Google Ads copy, the 90-character headlines don't leave a lot of room for required scope-of-services disclosures. That's a human judgment call every time.

Alberta (LSA). Rule 7.2 is more permissive than ON or BC on things like fee advertising, but the "demonstrably true, accurate, verifiable" test still applies. Alberta firms have a bit more runway on bold marketing, but that runway doesn't extend to unverified AI claims.

Quebec (Barreau). Rule 147 on its own isn't the whole picture. You've got the French-language content parity question, the trust-account disclosure requirements in fee ads, and a generally stricter posture on specialty claims. AI-generated content in Quebec needs a francophone lawyer's sign-off before it goes live. Period.

I'd also flag: CASL applies to your outreach even when the content itself is fine. If your AI workflow includes generating cold emails to prospective clients, CASL is going to be a bigger compliance risk than your provincial marketing rules. You need express consent. No workaround.

The Workflow That Keeps You Compliant

Here's what a defensible AI content process actually looks like, week by week, inside a firm. This is what I recommend to managing partners who want to use AI without the bar complaint risk.

Week 1: Set the policy. Before anyone generates anything, write down (a) which tools are approved, (b) what content types AI can draft versus what it can't touch, (c) who the human reviewer is for each content type, and (d) what goes in the record-keeping log. No policy means no defence when something slips through. For a template, see our piece on whether your firm needs an AI policy.

Week 2: Define the "never" list. This is the short list of content types AI is not allowed to draft at your firm. Mine usually includes: client testimonials or reviews of any kind, case result claims with specific dollar figures, success-rate statements, specialty claims, jurisdiction-specific legal advice in FAQ format, and any content going on the firm's French-language Quebec pages without a francophone lawyer in the loop.

Week 3: Build the drafting + verification flow. A typical flow looks like this. (1) Associate or marketing coordinator drafts content with AI assistance. (2) Every factual claim gets a source citation pulled manually, not by the AI. (3) A lawyer at the firm reviews for accuracy, jurisdiction fit, and marketing-rule compliance. (4) The reviewer signs off in writing (email is fine) before publication. (5) The draft, the prompt used, and the sign-off get logged somewhere retrievable.

Week 4: Test it on something low-stakes. Before you run this on a new cornerstone page, test it on a blog post. Run the full cycle. Find where it breaks. Fix the process. Then move to higher-stakes content.

Month 2 and ongoing: audit quarterly. Pull 5-10 random published pieces each quarter. Trace them back through your log. Did the review happen? Is the factual claim still accurate? Did anything drift? If yes, fix the piece and tighten the process.

The core point: the workflow isn't AI-specific. It's the same workflow you should already have for any content a non-lawyer at your firm produces. AI just lets more content get produced faster, which means your review bottleneck matters more.

For more on specific tools and how they handle citations and hallucinations, see our Lexis+ AI vs Westlaw Precision vs ChatGPT buyer guide and our breakdown of AI hallucinations in legal work.

A Quick Cost-of-Compliance Math Example

A lot of managing partners ask me if this verification overhead kills the ROI of AI-assisted content. Let's do the math.

Assume you're generating 4 blog posts a month. AI drafts each one in about 20 minutes. A lawyer's review, factual verification, and sign-off runs about 45 minutes per piece. That's 3 hours of lawyer time per month on AI content review.

Using DataForSEO's reported Canadian CPC of around CA$14.90 on "seo marketing for lawyers" and CA$39.65 on "law firm seo," the paid-traffic equivalent of a single well-ranked organic blog post (say, 200 monthly visits) is roughly 200 × $14.90 = $2,980 in equivalent ad value per month, per post. Four posts per month compounds from there.

3 hours of lawyer review at, say, $400/hour internal cost = $1,200/month. Four properly-ranking posts = ~$11,920/month in equivalent paid-search value. The math works, even with the compliance overhead baked in. What kills ROI isn't the review time. It's shipping AI content that doesn't rank, or worse, AI content that gets you flagged.

What AI Content Can and Can't Do For Your Firm

Based on pattern observations across the firms I've worked with, here's roughly where AI earns its keep and where it doesn't.

Where it works: first-draft blog posts on well-understood topics, outline generation, meta descriptions, internal linking suggestions, research synthesis for content you'll then rewrite, and FAQ drafts that a lawyer will edit before publication.

Where it doesn't: anything client-facing that makes a specific factual claim (case results, fee quotes, practice-specialty statements), French-language Quebec content without a francophone reviewer, testimonials of any kind, and advice-adjacent FAQ answers without a lawyer rewriting them in full.

When firms try to use AI to replace the lawyer's pen entirely, the content reads generic, doesn't rank, and creates compliance risk simultaneously. When firms use AI to compress a 4-hour writing session into a 1-hour editing session, they typically ship more content, better content, and stay on the right side of the Law Society.

For a related angle on this tradeoff, our piece on whether you should advertise AI-powered legal services goes into the brand positioning side of this question, which is separate from the content-compliance question I'm covering here.

Red Flags to Watch in Your Own Firm

If you're already using AI in your content workflow, here's a quick checklist to run against what's live on your site right now.

Any client testimonial, quote, or review attributed to a named or unnamed person. If yes, pull it and check if it's real. If you can't prove the person exists and said those words, take it down today.
Success rate claims, case value claims, "best" or "top" claims anywhere on your site. If yes, can you produce the records to back them? If not, rewrite.
Specialty claims ("our specialists in..." "our expert immigration team..."). If you're in a province that restricts these, rewrite.
French-language pages that were machine-translated and not reviewed by a francophone lawyer. Pull them until a human reviews.
Blog posts or FAQ pages with specific statute references, limitation periods, or procedural advice. Have a lawyer verify every one.
Google Ads copy without the required provincial disclaimer language. Fix today.
Google Business Profile, Analytics, or ad accounts in your agency's name rather than the firm's. This isn't an AI problem, but it's the single biggest operational risk I see. Get ownership transferred now.

If three or more of those come back as "uh oh," stop generating new AI content until you've fixed the process. The bar complaint risk isn't worth four more blog posts this month.