No Ontario or federal law requires a real estate brokerage to hand a client a written “how we use AI” policy. But CREA’s own position statement says clients should be able to tell when AI is materially shaping what they’re being told, and the closest Canadian professional framework for deciding when disclosure is actually owed comes, of all places, from the legal profession.
Key takeaways
CREA’s national position on artificial intelligence states plainly: “The use of AI must be guided by transparency, accuracy and accountability,” and that “consumers and REALTORS® should clearly understand when AI is materially influencing information, communications, or representations.” CREA, position on artificial intelligence The page ties that position to the REALTOR® Code’s existing advertising articles, Article 13 (Content/Accuracy) and Article 15 (Advertising Claims) — and Article 15.6 of the Code itself requires that “any claims or offerings in advertising must also comply with all applicable laws, including the Competition Act.” REALTOR® Code, Article 15.6 What the page does not say is just as important for accuracy’s sake: it sets out no specific informed-consent standard, no employee-count threshold, and no required disclosure script. Those simply aren’t on the page, whatever secondary summaries might claim.
The federal government’s Voluntary Code of Conduct on advanced generative AI lists its 46 signatories by name, and one of them is the Organisme d’autoréglementation du courtage immobilier du Québec — OACIQ, Quebec’s own real estate brokerage regulator. ISED, Voluntary Code of Conduct, signatory list Among the Code’s measures is a manager-level obligation that systems “that could be mistaken for humans are clearly and prominently identified as AI systems.” That obligation binds OACIQ as an organization managing AI systems, not individual Quebec brokers, and the Code itself remains voluntary — but it’s a real, sourced example of a Canadian real estate regulator engaging directly with this question rather than staying silent on it.
The Law Society of Alberta’s survey of Canadian generative-AI guidance records the Law Society of Ontario Futures Committee’s four-factor test for when a lawyer should tell a client AI was used, verbatim: “Will the use of Gen AI necessarily be disclosed publicly…? Does the client reasonably expect that the material being prepared by Gen AI would actually be prepared by a licensee? Are there reputational or other forms of risk to the client that could arise from the use of Gen AI? Does use of Gen AI require inputting of the client’s personal or proprietary information?” Law Society of Alberta, Gen AI Rules of Engagement This is a legal-profession test, reasoned here purely by analogy — nothing requires a real estate agent to apply it. But the four questions translate cleanly: would a client be surprised, is there a reputational cost if it goes wrong, and does answering the question require handing the tool any of the client’s own information, which loops straight back into the confidentiality question covered elsewhere on this hub.
RECO’s confidentiality bulletin already publishes a consent-disclosure template built for a different purpose — sharing a client’s confidential information — but it repurposes cleanly into a policy skeleton: what information will be shared, who receives it, what the purpose is, and who benefits and how. RECO Bulletin 2.5, Confidentiality Applied to an AI policy, that becomes: which tools the practice uses, what client information (if any) goes into each one, why, and what the client gets out of it. Reusing a real, sourced Ontario structure is a more defensible starting point than inventing a template from nothing.
A worked example applying the four-factor test: an agent uses one AI tool to draft a comparative market analysis narrative from MLS data the client already knows is public, and a separate tool to summarize private phone notes about the client’s moving timeline. The CMA narrative doesn’t require the client’s private information and carries little reputational risk if wrong, since it gets reviewed before it goes out — on the four-factor logic, that one probably doesn’t need a specific disclosure. The call-notes summary does use the client’s own private information, which is exactly the fourth factor and exactly the kind of disclosure RECO’s own consent structure is built around — that’s the one worth naming plainly in a policy or asking consent for.
Two separate pressures point the same direction even though neither is a binding legal requirement on its own. CREA’s own transparency-accuracy-accountability position is a national professional-association statement, and the Ontario legal profession’s four-factor test is reasoned here by pure analogy — neither is a statute a real estate regulator will enforce against a brokerage today. But together they describe the same client-facing risk from two different professions: a client who finds out later that AI shaped what they were told, without ever having been told that, is a client whose trust the transaction may not survive. A short, honest policy costs little to write and answers the question before a client has to ask it.
Keeping the policy narrow and specific also avoids overpromising. A policy that simply says “we may use AI tools to help draft materials, and we review everything before it reaches you” is both accurate and enough — it doesn’t need to list every tool by name or make a claim about accuracy the practice can’t actually back up. The review-step discipline covered elsewhere on this hub is what makes that second half of the sentence true.
Related: keeping client details out of a public model, the review step that keeps you out of trouble, AI for listing and social marketing.
Not a specific one. CREA’s national position asks that consumers be able to tell when AI is materially shaping information they receive, but that’s a professional-association statement, not a statute, and no Ontario or federal law sets out a required disclosure script for real estate specifically.
The Law Society of Ontario’s four-factor test — built for lawyers, applied here by analogy — is a reasonable one: whether disclosure will happen anyway, whether the client would expect a person did the work, whether there’s reputational risk if it’s wrong, and whether the client’s own information had to go into the tool.
Yes — Quebec’s OACIQ is a listed signatory to the federal government’s Voluntary Code of Conduct on generative AI, which commits signatories managing AI systems to identify them clearly when they could be mistaken for a human. It’s voluntary, and it binds OACIQ as an organization rather than individual Quebec brokers, but it’s a real, sourced example of a Canadian real estate regulator taking a position.
A short call is enough to turn this into a one-page policy you can hand a client without guessing.
