Attaching the word “AI” to a claim does not move it into a different legal category. The Competition Act asks the same questions of an AI-related claim it asks of any other — and one provision in particular puts the burden of proof on whoever is making the claim, not on a regulator to disprove it.
Key takeaways
A business does not get a different set of advertising rules because a language model, rather than a person, chose the words. The Competition Act regulates the representation, and it has never mattered, for that purpose, who or what drafted it.
The Competition Bureau explains the standard directly: “When deciding whether a representation is false or misleading, a court is required to take into account the general impression conveyed by the representation, in addition to its literal meaning.” (Competition Bureau, the general impression test) That test governs section 52 and paragraph 74.01(1)(a) equally, and it does not ask who authored the copy — a marketing team, a freelance writer, or a generative AI tool drafting the first pass of an ad.
The Act’s language on performance claims is specific and, for AI-related marketing, easy to trip over. Paragraph 74.01(1)(b) prohibits a representation “in the form of a statement, warranty or guarantee of the performance, efficacy or length of life of a product that is not based on an adequate and proper test thereof, the proof of which lies on the person making the representation.” (Competition Act, s.74.01)
That last clause is the one worth reading twice: the proof sits on the business making the claim. If an AI-powered feature is marketed as detecting fraud, drafting better emails, or cutting missed payments by some margin, that is a performance claim the moment it is published — and the business has to be able to show the adequate and proper test existed before the claim went out, not scramble to produce one after a regulator asks.
The exposure behind that reversed onus is not abstract. Where a court finds reviewable conduct under section 74.01, it may order an administrative monetary penalty against a corporation of up to the greater of two figures: $10,000,000, rising to $15,000,000 for a subsequent order, and three times the value of the benefit derived from the conduct or, where that benefit cannot be reasonably determined, 3% of the corporation’s annual worldwide gross revenues. Competition Act, s.74.1 See Treadstone Law’s guide to misleading advertising under the Competition Act for how a civil reviewable-conduct file actually proceeds.
Dynamic or AI-personalised pricing does not sit outside the Act’s pricing rules. Subsection 74.01(1.1) states, “the making of a representation of a price that is not attainable due to fixed obligatory charges or fees constitutes a false or misleading representation,” with a narrow carve-out for charges imposed by law. (Competition Act, s.74.01(1.1)) Whether a human set the headline price or a pricing model adjusted it in real time, the price a customer is actually shown to pay has to be the price they can actually pay.
A separate provision reaches specifically into the mechanics of an electronic message. Section 52.01 prohibits a false or misleading representation “in the sender information or subject matter information” of a message, in the body of the message itself, and in a locator; subsection (4) removes any doubt about intent, stating “it is not necessary to prove that any person was deceived or misled.” (Competition Act, s.52.01) An AI copywriting or personalisation tool drafting a subject line or a body paragraph does not change what that provision is checking for.
A generative copywriting tool asked to make a product sound better than the competition will often reach for an environmental or comparative claim, because those read as concrete and persuasive. The Act treats those the same way it treats a performance claim: paragraphs 74.01(1)(b.1) and (b.2) prohibit environmental-benefit claims about a product, or about a business or business activity, that are not based on “an adequate and proper test” or “adequate and proper substantiation” — the same reversed-onus structure as the performance-claim rule above. (Competition Bureau, misleading representations and deceptive marketing practices) An AI tool does not know whether the business behind it actually has the substantiation on file — it will write the claim regardless.
A business that genuinely wants to know where a specific AI-related claim stands before it publishes does not have to guess. The Bureau offers written opinions under section 124.1 of the Act: contact the Information Centre, and “a written opinion is binding on the Commissioner as long as the facts submitted are accurate,” remaining binding as long as those facts stay substantially unchanged and the conduct is carried out as proposed. (Competition Bureau, use of tests or testimonials) A fee applies, set out in the Bureau’s own fee schedule, but for a claim a business intends to run at scale, that is a materially cheaper way to get certainty than finding out during an investigation whether the underlying test was adequate.
A brokerage adopts an AI copywriting tool that drafts an ad claiming its process “reduces missed payments by 40%.” Nobody at the business tested that figure — the tool generated a plausible-sounding number to make the copy punchier. Under paragraph 74.01(1)(b), that is exactly the claim the reversed onus is built for: if a regulator asks, the business has to produce the adequate and proper test behind the 40% figure. “The AI tool wrote it” is not a defence, because the Act attaches responsibility to the person promoting the product, not to the process that generated the wording.
Related: what “AI washing” means, AI-generated reviews and testimonials, and CASL and AI-written email.
How a business builds AI-assisted marketing with this compliance layer already accounted for is covered on the AI growth and marketing hub.
Review does not create an exemption. Paragraph 74.01(1)(a) covers a representation made “by any means whatever,” and responsibility sits with the business promoting the product or business interest, regardless of who or what produced the specific wording, or whether someone signed off on it afterward.
Not under the Act. Liability attaches to the person making or causing the representation to be made, and section 74.01 does not carve out an exception for wording a business obtained from a vendor’s AI feature rather than writing itself.
No differently at all — paragraphs 74.01(1)(b.1) and (b.2) apply the identical reversed-onus test to environmental claims that paragraph 74.01(1)(b) applies to performance claims, regardless of what produced the wording.
A short call is enough to walk through a specific claim against the reversed-onus rule above.
