Anonymised, illustrative composite. A brokerage assistant used a general-purpose AI writer to turn rough notes into MLS-ready copy. The tool filled a gap in the notes with a plausible-sounding fact that was not true.
At a glance
A listing coordinator fed rough seller notes into a general-purpose AI writing tool to speed up MLS copy for a Hamilton detached home. The notes said the kitchen had “updated cabinets, newer counters” — nothing about appliances or a renovation year. The tool's output, pasted into MLS with only a skim for typos, read: “fully renovated kitchen with new appliances (2024).” Nobody on the file had told it 2024, or new appliances, or the word “fully.” It filled the gap the notes left open with something that sounded right, in the same confident register as everything true around it — which is exactly why nobody caught it on the read-through.
A buyer's home inspector noted the range and dishwasher showed wear consistent with roughly a decade of use, not a 2024 replacement, and flagged the mismatch against the listing description in his report. The buyer's lawyer raised it during the financing condition period, not after closing — which mattered, because it gave the buyer a live off-ramp rather than a lawsuit. A second, smaller gap turned up in the same review: the same AI pass had also softened “newer counters” into “quartz countertops,” and the counters were laminate. Neither invented fact was dramatic on its own; together they read as a pattern of upgrades that had not happened.
Nothing here turns on an invented dollar figure, deliberately — the point of this case is that the AI tool invented a fact, not a number, and the fix does not require costing out a repair that was never needed in the first place. The only figure worth naming is procedural: the financing condition on this file ran 10 business days, which was enough time for the inspection to surface the gap and for the description to be corrected before the buyer had to decide whether to waive. A firm-offer structure with no financing condition would have removed that window entirely.
Canada does not treat marketing claims as consequence-free once a human stops typing them. The Competition Act, s.52, prohibits making a representation to the public that is “false or misleading in a material respect,” knowingly or recklessly — and specifically does not require proof that anyone was actually deceived, or that the public reached was located in Canada. Section 74.01(1)(a) mirrors that on the civil-reviewable track, which is the more realistic exposure for a single listing description than a criminal referral under s.52 would be. Neither section carves out an exception for who, or what, drafted the sentence — the representation is the brokerage's and the agent's the moment it is published under their name.
The Canadian Real Estate Association addresses this directly for members: “REALTORS® must remain fully accountable for the information, advice and services they provide to clients,” pointing to REALTOR® Code Article 13 (Advertising: Content and Accuracy) and Article 15 (Advertising Claims). CREA is explicit that the standard travels with the person publishing the content, not the tool that generated a draft of it, and that “accuracy, clarity, and transparency remain essential to informed decision-making and maintaining confidence in the real estate transaction process.”
The brokerage corrected the MLS description before the financing condition expired, disclosed the correction to the buyer's lawyer in writing, and the deal closed on the original terms with the buyer satisfied the price already reflected the kitchen's real condition. Had the mismatch surfaced after an unconditional close instead, the same two facts — a renovation year and a countertop material, neither true — would have supported a materially harder claim: a representation the buyer could show they relied on, made about a condition that turned out to be false, with no financing condition left to walk through.
The failure was not the AI tool; it was publishing its output with no fact-check against the actual seller notes. An AI draft is not evidence of anything — it is a draft, exactly like one written by a junior copywriter would be, and it needs the same read-against-the-source pass before it goes to MLS, line by line, not a skim for typos. CREA's own framing is useful here: the transparency question is not whether AI touched the copy, it is whether “consumers and REALTORS® clearly understand when AI is materially influencing information, communications, or representations” — and the fastest way to fail that test is to never have checked what it wrote. A one-line house rule closes most of the gap: nothing an AI tool adds that is not already in the seller's own notes goes to MLS without a named person confirming it against the file.
A 30-minute call is enough to tell you whether AI pays for itself here.
