FR
← Blog
AI and trust

Disclosing AI-generated content in advertising: what is required, and by whom

September 2026 · AI and trust · Québec

For an ordinary commercial ad, neither Meta nor Google shows a rule, in the policy pages reviewed, that requires you to disclose AI-generated content. Their obligations cover election, political and social-issue ads. That does not mean anything goes. Canada's Competition Act applies to every ad you run, and courts weigh the general impression it creates. So what you must declare comes down to one question: will the buyer take this generated element as a fact about your company? If the answer is yes, a label often is not enough. The element has to be replaced.

This page covers a single moment: the generated content is already in the ad, and you have to decide what to say about it. It describes public rules and is not legal advice.

What Meta and Google actually require

Start with what the platforms write themselves, because this is where the most unsourced claims circulate.

At Meta, the rule sits in the policy on ads about social issues, elections or politics. Advertisers must disclose a photorealistic image or video, or realistic-sounding audio, created or edited with third-party generative AI tools. Examples include a real person shown saying or doing something they did not, a realistic person or event that does not exist, or altered real footage. Without the disclosure, Meta rejects the ad, and repeated failures can lead to penalties against the advertiser. Meta adds that beginning June 1, 2026, it also uses automated detection technology to identify ad media created or edited with those tools.

At Google Ads, the equivalent rule lives in the political content policy and covers election ads. Synthetic or digitally altered content that inauthentically depicts real or realistic-looking people or events must be disclosed. Edits that are inconsequential to the claims made in the ad are exempt: resizing, cropping, colour or brightening corrections, defect correction. Depending on the format, the disclosure is generated automatically or the advertiser must add a prominent one.

Here is the point that matters to an industrial firm. In the pages reviewed, neither Meta nor Google shows an AI disclosure rule for ordinary commercial advertising. Meta's general Advertising Standards page contains no such language. A parts distributor in the Montérégie promoting its product line on Facebook therefore has, according to those pages, no platform obligation on this point. Be wary of a supplier who tells you otherwise without showing you the page. Be just as wary of one who concludes that nothing applies.

Organic content has its own rules

Unpaid posts follow a different regime. Meta requires its AI disclosure tool whenever someone posts organic content with photorealistic video or realistic-sounding audio that was digitally created or altered, and it may apply penalties when they do not.

YouTube requires disclosure when content makes a real person appear to say or do something they did not, alters footage of a real event or place, or generates a realistic scene that did not happen. YouTube does not require it for caption creation or for production assistance, such as an outline, a script, a thumbnail or a title prepared with a generative tool. Creators who consistently choose not to disclose may face a label applied manually, removal of content or suspension from the YouTube Partner Program. If the same video runs as an ad and as a post on your channel, check both regimes, not only the one for ads.

The rule that covers every ad: general impression

The Competition Bureau of Canada pages reviewed do not mention AI. They do not need to. Under the Competition Act, it is against the law to advertise or market something in a way that is false or misleading. Courts consider the general impression a representation conveys, in addition to its literal meaning. And the Bureau states that no one actually needs to be deceived for a practice to be found misleading.

Three of its tips for advertisers bear directly on generated content. The Bureau does not tie them to AI; applying them this way is our reading.

The third point is the one people forget. A small "AI-generated image" line under a plant that does not exist does not correct the impression that it is your plant. It contradicts it, which is exactly the job a disclaimer must not do. Disclosure does not repair a misleading image. It helps when the generated element is already honest in its context and you want to remove any doubt. The same reasoning about general impression and fine print is set out on the page about client testimonial advertising rules.

What the client relationship requires, rule or no rule

A buyer at a prime contractor does not read Meta's policies. They read your ad, then they come to visit the plant. That is the day the disclosure question is really settled. The useful test fits in one sentence: if the buyer found out later that this element was generated, would they feel misled?

Applied to real cases, the test sorts quickly.

The logic is the same as with stock images: where the buyer is looking for proof, an image that is not yours costs credibility, label or not. The page on stock images vs real photos draws that line for the website. In advertising it costs more, because you are paying to distribute the image.

What to demand from your supplier comes down to three things, in writing. A list of the elements generated or altered with AI in each ad, image by image and audio track by audio track. Confirmation that every image showing your plant, your products or your people was captured on your premises. And, for any ad on a social or political issue, proof that the disclosure the platform requires was made. The answer that should end the conversation: a supplier who cannot tell you what, in your own ad, was generated.

On our side, the person who shoots is also the person who runs the ad accounts, so the origin of every image in an ad can be traced without going through an intermediary. That is the idea behind "Certified Human Content": what serves as proof is captured, not generated. AI tools keep a useful place where they do not touch the proof, such as a first draft of subtitles or the transcript of an interview, reviewed by a person before publication. How the campaigns themselves are run is described on the Facebook and Meta advertising page.

When this page does not apply to you

If your ads contain no generated element, there is nothing to declare. A normal edit, a crop or a colour correction that does not change what the image shows, does not raise the question. Google explicitly exempts such edits in its election ad rule.

If you run political or social-issue advertising, for example a campaign on a regional project that divides opinion, this page is not enough. Each platform's own regime applies in full, with its own requirements, and that is a conversation to have with your counsel before launch.

Finally, if AI was used only for production assistance, an outline, a first script draft, a title, the disclosure question does not arise in the same terms. YouTube says as much for its own videos. For the buyer, what matters is that every claim in the ad is true, not which tool helped phrase it.

The policies cited here change; Meta's automated detection, for instance, dates from June 2026. Reread the platform's page when you launch a sensitive campaign. This page is general information, not legal advice.

Sources

If your ads carry images whose origin you cannot trace, book a short call and we will see whether we can be useful to you.

Book my call
Home · Portfolio

B2B Marketing · Industrial Marketing · Uncompressed · ClientGallery

© 2026 MarketingSTRAT