Nothing in Canadian law singles out an AI-generated image for special treatment. What that actually means is less reassuring than it sounds: every ordinary rule that already applied to marketing imagery still applies, in full, to the AI-generated kind.
Key takeaways
Start with what usually is not the problem. Whether an AI-generated image itself qualifies for copyright protection is a live and separate question, covered in a related article on this hub, but it is rarely the risk that actually bites a business using AI images in its own marketing. The more common exposure sits in advertising law: what the image tells a customer, truthfully or otherwise, about the product being sold.
Section 74.01(1)(a) of the Competition Act makes it reviewable conduct to make a representation to the public that is false or misleading in a material respect, for the purpose of promoting a product or business interest, by any means whatever. An AI-generated image that shows a product looking different from how it actually looks — a different colour, a feature it does not have, a scale or finish that misrepresents the real item — falls squarely inside this provision. The statute does not ask whether a camera or a generative model produced the image; it asks whether the representation to the public is accurate.
The same section goes further for performance claims specifically: a representation in the form of a warranty or guarantee of a product’s performance, efficacy or length of life is reviewable unless it is based on an adequate and proper test, and the burden of proving that test was done sits on the business making the claim. An AI-generated “action shot” that implies a performance capability the product has never actually demonstrated carries this same reversed-onus risk.
C2PA describes itself as an open technical standard for establishing the origin and edits of digital content, delivered as what it calls Content Credentials — described on its own site as functioning “like a nutrition label for digital content”. Using a tool like this to label AI-generated marketing images is a legitimate transparency practice that some businesses are adopting voluntarily. It is worth being precise that no Canadian statute currently requires this kind of labelling for commercial marketing images — adopting it is a choice about trust with an audience, not compliance with a specific legal duty.
treadstonelaw.ca’s overview of Ontario advertising and marketing law covers the general terrain that an AI-generated visual can just as easily walk into: rules on endorsements and testimonials, contest and promotion mechanics, and pricing claims shown alongside an image. An AI-generated image depicting a person appearing to endorse a product, for instance, raises the same endorsement-disclosure questions a staged photograph of a paid actor would — the medium used to produce the image does not change which advertising rules apply to the claim it makes.
Where the image depicts a real person, a separate right can apply on top of everything above. Quebec’s Civil Code lists, among the acts that may be considered an invasion of privacy, “using his name, image, likeness or voice for a purpose other than the legitimate information of the public” (Civil Code of Québec, art. 36). An AI-generated image built to look like a real, identifiable individual endorsing a product — without that person’s consent — can therefore raise a privacy claim distinct from, and in addition to, the misrepresentation exposure above.
A furniture retailer generates a lifestyle image showing a sofa in a fabric colour the retailer does not actually stock, because the generated image rendered more appealingly than the real sample photos. Publishing it as-is risks a Section 74.01(1)(a) misrepresentation claim the moment a customer orders based on a colour that was never available — not because the image was AI-generated, but because it makes a false representation about the product on offer. The fix has nothing to do with the generation tool: match the image to what is actually sold, or clearly caption it as a styled rendering rather than the literal product, before the campaign goes live rather than after the first order comes in.
The advertising-law risk above is the one that most often bites, but it does not replace the copyright question — it sits alongside it. If the image was produced by a tool trained on other people’s photography or artwork, the same fair-dealing analysis that governs AI training generally can be relevant on the input side even where the output-side ownership question stays unresolved. The two questions — can you be sued over what went into the image, and do you own what came out of it — are genuinely separate, and clearing one does not clear the other.
A misleading product image carries different specific consequences depending on where it runs — a website claim, a paid social ad, an email campaign — but treadstonelaw.ca’s survey of Ontario advertising rules shows the underlying principle repeating across each of them: what is being represented to a consumer, and whether that representation is accurate, is the constant question regardless of the medium or how the creative was produced. A business rolling out AI-generated imagery across several channels at once should check each placement against that same question rather than assuming clearance on one channel covers all of them.
No law bans it outright, and no Canadian statute specifically singles out an AI-generated image for a heavier standard than a photograph. The exposure comes from what the image represents about the product, under the same misleading-advertising rules that apply to any marketing image, not from the fact that it was AI-generated.
There is no general Canadian statute requiring that disclosure for commercial marketing images. Tools like Content Credentials make voluntary disclosure practical for businesses that want to adopt it as a transparency practice, and some industries or platforms may layer their own disclosure expectations on top of the general legal baseline.
That would run through ordinary copyright or passing-off analysis depending on the facts — whether protected expression was actually copied, or whether the image creates confusion with a competitor's brand — the same tests that would apply to a photograph raising the same concerns.
It can help on the transparency side and may reduce the chance a viewer is misled about how the image was produced, but it does not cure a false representation about the product itself — a clearly labelled AI image that still misstates a product's colour or features remains a misrepresentation under Section 74.01(1)(a) regardless of the label.
Related: AI, your brand, and trademarks and whether training an image model on copyrighted work infringes copyright.
Checking what an AI-generated image actually represents about a product is part of scoping a growth-marketing engagement properly.
