Dental marketing

AI Content for Dentists: What RCDSO, CDSBC, and CDSS Actually Allow

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

Most dental marketing guides opened 2026 by telling you ChatGPT would replace your copywriter. Ignore that. The real question isn't whether AI content for dentists works, it's whether your provincial college will let you publish the thing you just asked an AI to write.

Here's the thing. Canadian dental colleges have specific advertising rules that predate AI by decades. Those rules still apply whether a human or a language model wrote the sentence. And a lot of the AI-generated dental content I've audited this year, blog posts, service pages, social captions, would get a practice in trouble under Ontario Regulation 853/93 or the CDSBC Code of Ethics if anyone actually read it.

So this article is about the intersection. What can you safely use AI for in your practice marketing? What will trigger a college complaint? And where's the line between "helpful draft" and "regulatory problem"?

I'll focus on Ontario (RCDSO), British Columbia (CDSBC), and Saskatchewan (CDSS), because those are the three provinces I get asked about most. Alberta (ADA&C) and Quebec (ODQ) have similar structures but different specifics, check your own college's advisories. If you want a broader breakdown of how SEO and content fit together for a practice, that's covered in our complete dentist SEO guide. This piece is narrower, it's the compliance layer underneath all of it.

The Core Rule Every Canadian Dental College Agrees On

Before we get to AI specifics, you need to understand what all the colleges share in common. Because the AI question basically boils down to "does the AI help me violate this, or help me comply with this?"

Three rules show up in every provincial framework:

1. No testimonials. Under the RCDSO Advertising Guidelines, you cannot publish "testimonials or any statement that can only be verified by a person's personal feelings or views." CDSBC and CDSS have near-identical language. This matters a lot with AI, because if you ask ChatGPT to "write a patient success story for my Invisalign page," it will happily produce something that reads like a testimonial. That's the violation. The AI doesn't know the rule.

2. No superlatives or claims of uniqueness. Per the RCDSO, you cannot claim to be "#1 rated," "cutting edge," "the best," or suggest superiority over other practices. AI-generated copy defaults to this language constantly. Every time I paste a blog draft from a default ChatGPT prompt, I find at least two superlative claims my editor has to strip out.

3. Truthful, not misleading. All three colleges require advertising to be "legal, decent, honest and truthful" (RCDSO's wording, and the others echo it). AI hallucinates. It will invent statistics, fabricate credentials, and confidently state treatment outcomes that aren't supported by evidence. If you publish an AI-written blog that claims "studies show implants last 40 years on average," and that's not actually what studies show, you've published misleading content. That's your problem, not the AI's.

Keep those three rules in your head for the rest of the article. Everything else flows from them.

What AI Content Is Genuinely Safe For

Not everything an AI writes is a compliance land mine. In my experience, practices get in trouble when they use AI to generate patient-facing claims, and they benefit when they use it to generate patient-facing information.

Here's where AI actually shines, and where I'd use it without losing sleep:

Educational content about procedures. A blog post explaining "what happens during a root canal" is factual, widely documented, and doesn't require testimonials or claims of superiority. AI writes these in a few minutes. You still need a clinician to fact-check it, but the first draft is legitimate use. Per Statistics Canada's 2024 data on health information searching, over 70% of Canadians look up symptoms or procedures online before booking, if you're not publishing educational content, you're invisible to them.

FAQ pages and appointment-prep instructions. "What to expect before your appointment," "how to prepare for a cleaning," "what we need from new patients." This is operational content. No claims, no testimonials, no comparisons to other practices. AI handles it well.

Internal drafts of policies and scripts. Staff training materials, intake call scripts, recall templates. Never patient-facing, so the college rules don't even apply. Just make sure anything you deploy publicly is reviewed by a human who knows the rules.

Meta descriptions and page titles. Short, technical, and usually fine, as long as you manually remove any "best dentist in Calgary" language the model wants to insert. Which it will. Every time.

Local SEO blog posts about your city or neighbourhood. "New patient guide to downtown Saskatoon" or "what to expect at a dental office in Vancouver" are mostly informational. Safe territory. For how this content fits into your broader strategy, see our complete guide to dental marketing in Canada.

Where AI Content Becomes a Compliance Problem

Now the parts that actually cause trouble. These are the categories where I've seen practices get letters from their college or have to take content down.

Service pages that drift into claims. A default AI prompt for "write a page about dental implants at my practice" will produce phrases like "our implants are long-lasting and natural-looking," "patients love the results," or "we provide the highest standard of care." Every one of those sentences is a problem under RCDSO rules. "Highest standard of care" implies superiority. "Patients love" is a testimonial in disguise. The AI doesn't know it's crossing a line.

Before-and-after galleries with AI-written captions. The photos themselves require explicit written consent per RCDSO (and all three colleges agree). But the AI-generated captions, "amazing change," "incredible results", add a second violation layer. Even if the photo is compliant, the caption isn't.

AI-generated "patient stories" or "case studies." This one trips up a lot of practices. They know they can't publish real testimonials, so they ask ChatGPT to write a "representative patient journey." The college doesn't care that the patient is fictional, if it reads like a testimonial and influences patient decisions, it's still prohibited promotional content. Under the RCDSO guidelines, content that "creates an expectation of favourable results" is restricted regardless of whether the patient is real.

Superlative-heavy social captions. Instagram is where this breaks most often. AI caption generators love words like "stunning," "life-changing," "best in the city." If you're posting to a patient-facing account, those captions need human review before they go live. For the full picture on platform-specific rules, see our dental social media marketing guide.

Fabricated statistics. The one that scares me most. AI models will confidently invent a percentage, a study, or a survey result that sounds plausible but doesn't exist. If you publish "87% of our patients report a brighter smile within two weeks" and that number came from a hallucination, you've violated the "truthful and not misleading" requirement. Every AI-generated stat needs a real source, or you cut it.

A Practical Workflow for Using AI Without Getting Burned

Here's the workflow I actually recommend to dental practices that want to use AI content without ending up in a college complaint. This is the week-by-week version.

Week 1, build your prompt library. Don't start generating. Start defining what the AI is allowed to do. Write three to five standard prompts for your common content needs: educational blog posts, FAQ entries, procedure explanations. In each prompt, explicitly tell the AI: "Do not use superlatives. Do not include testimonials or patient stories. Do not claim superiority over other practices. Do not invent statistics, only include data if I provide it." This single step eliminates maybe 60% of compliance risk before you ever generate a word.

Week 2, build your review checklist. Every piece of AI content that goes public gets reviewed against a standard checklist. Mine has seven items: superlatives stripped, testimonials removed, claims sourced, stats verified, treatment outcomes softened to "may" or "can" rather than "will," all clinical language checked by a dentist, final read for tone. The checklist is the firewall between AI draft and published content.

Week 3, pilot with low-stakes content. Pick the safest category, usually educational blog posts or FAQ entries, and run your workflow end-to-end on two or three pieces. Publish them. See what your team catches during review. Refine the prompt library based on what went wrong.

Week 4, expand cautiously. Add one more content category, maybe service pages or recall emails. Do not expand to social media captions or case studies until your workflow is solid on the easier stuff. Most practices I see rush this step and regret it.

Month 2 and beyond. Audit quarterly. Pull ten pieces of AI-assisted content off your site and ask: would this survive a college complaint? If you can't say yes for every single piece, you have a gap in the workflow.

Typically, practices that skip the prompt library step end up with content that reads generic and off-brand. Most practices that skip the review checklist end up publishing at least one compliance violation within the first six months. The workflow exists because both of those failures are common.

The Cost Math: Is AI Content Actually Cheaper?

Everybody's selling AI as the cheap option. Let me show you the honest math.

Assume you publish two blog posts per month, 1,200 words each. A Canadian dental content writer bills somewhere between $0.50 and $1.50 per word, call it $1.00 at the midpoint based on typical freelance rates in the dental space. That's $1,200 per post, or $2,400/month for two posts. Round numbers.

With AI assistance, the draft generation is effectively free (ChatGPT Plus is $25 CAD/month, Claude Pro similar). But you still need a human to do the compliance review, fact-check clinical claims, and edit for voice. In my experience, that's about 60 to 90 minutes per post for a decent editor billing $75-$120/hour.

So the real AI-assisted cost is something like: 75 minutes × $100/hour = $125 per post, plus the $25/mo tool cost. Two posts = $275/month. Versus $2,400/month for full human-written.

That's a real saving. But check your actual numbers, because if your editor isn't actually stripping superlatives and checking claims, the "saving" is debt you'll pay in a college complaint down the road.

The per-new-patient math matters too. General dental practice cost-per-new-patient benchmarks sit in the $150-$400 range (industry standard, check your PMS software for your actual number). If AI-assisted content helps you publish more frequently and capture more organic traffic, the savings compound. If AI-assisted content gets flagged by your college and you have to pull pages, the costs compound the other way. The workflow determines which side you land on.

Provincial Specifics You Actually Need to Know

Three quick province-level notes, because the colleges don't word-for-word agree on everything.

Ontario (RCDSO). The advertising guidelines are the most detailed of the three. Beyond the core rules, they explicitly address social media and require dentists to be responsible for all content involving them or their clinic, including staff posts and third-party content. That means if you hire an AI tool that auto-generates social posts, you're responsible for every one. Read the RCDSO Advertising Guidelines directly, they're publicly posted on rcdso.org.

British Columbia (CDSBC). The CDSBC Code of Ethics and Professional Conduct covers advertising under similar principles but puts more emphasis on informed consent for any patient-related content. If you're in BC and considering AI-generated patient education materials, the standard is whether the content would reasonably inform rather than persuade.

Saskatchewan (CDSS). CDSS follows similar themes with its own bylaws and advertising standards. Smaller province, fewer published advisories, but the core "no testimonials, no superlatives, truthful" framework applies. When in doubt in Saskatchewan, call the college. They're responsive and would rather give you guidance up front than deal with a complaint after.

For Alberta and Quebec, ADA&C and ODQ have their own rules with similar bones but important differences (ODQ rules interact with Bill 96 French-language requirements). Don't assume Ontario rules translate.

Decision Framework: Should You Use AI for This Piece of Content?

Before you generate anything, run the piece through this filter:

If the content is purely educational (procedure explanations, FAQ, prep instructions): AI-assisted is fine with a human review pass. Go ahead.

If the content is a service page describing what you offer: AI can draft, but a clinician and a marketing reviewer both need to strip claims, verify outcomes language, and remove superlatives before it publishes.

If the content involves any patient experience, review, or outcome story: Do not use AI. The risk of producing something that reads as a testimonial is too high, and the violation is clear-cut under all three colleges.

If the content includes clinical statistics or treatment outcome data: Only use AI if you're providing the stats yourself from verified sources. Never let the AI generate numbers.

If the content is a social media caption or ad copy: AI can help, but assume every draft needs human editing to remove superlatives and soften claims. The default AI tone is too promotional for compliance.

If the content involves before-and-after photos or visual patient results: The photo compliance matters more than the AI compliance. Handle the consent piece correctly first, then worry about the caption.

The rule that sits underneath all of this: AI is a draft tool, not a publish tool. If your workflow has an AI-generated piece going live without a human compliance review, you've built a liability machine. If your workflow has AI speeding up drafts that humans still edit and approve, you've built something useful.