AI ads and the EU AI Act: what you have to disclose

As AI-generated video moves from novelty to default, a second question follows the creative one: what do you have to tell people about how it was made? For brands operating in or selling to the EU, that question now has a regulatory edge. The EU AI Act introduces transparency obligations for artificial content, and they land squarely on the kind of work marketing teams produce.

This is not a reason to slow down, and it is not legal advice. Every brand should check specifics with its own counsel. But the shape of the obligation is clear enough that you can design for it now, and the teams that do will find it is far less disruptive than the headlines suggest.

Why this lands on marketing's desk

For years, the question of whether content was "real" was an ethics conversation, not a compliance one. The AI Act changes that for a specific category: content that is generated or manipulated by AI in a way that could be mistaken for authentic. Marketing video sits right in that category, because the whole point of good AI UGC is that it reads as genuine.

That is the tension the law is responding to. The more convincing your generated content is, the more the transparency obligation applies to you. So this is not a niche legal footnote that compliance handles in the background. It is a creative constraint that shapes how the asset is made and delivered, which makes it marketing's problem to solve well.

What the Act actually requires

The relevant provision is Article 50, the AI Act's transparency rules. In plain terms, two ideas matter most for marketers:

• Disclosure of artificial content. Where AI is used to generate or manipulate image, audio, or video content that resembles real people, places, or events (the "deepfake" category), deployers must disclose that the content has been artificially generated or manipulated. The audience should be able to know they are not looking at a captured recording of something that happened.
• Machine-readable marking. Providers of generative systems are expected to mark synthetic output in a machine-readable format so that it can be detected as AI-generated, supporting provenance standards rather than relying on a visible label alone.

The Act entered into force in 2024, and these transparency obligations apply on a staged timeline running into 2026, which is exactly why now is the moment to build the habit rather than retrofit it. There are nuances about artistic and editorial context, and the precise wording of a disclosure is not dictated, which leaves room to do it gracefully.

The law does not require you to make AI content look worse. It requires you to be honest that it is AI content. Those are very different design problems, and only one of them is hard.

Disclosure done well is a trust asset

Here is the reframe worth internalising. A disclosure obligation sounds like a cost: another label, another disclaimer, another thing that makes the ad feel less slick. Treated reactively, that is exactly how it will feel and look.

Treated as a design choice, it is the opposite. Audiences are already suspicious of synthetic media, and that suspicion is rising faster than any regulation. A brand that labels its AI content clearly, early, and in its own voice signals confidence rather than concealment. The brands that will be damaged are not the ones that disclosed; they are the ones that got caught not disclosing. Clear provenance is becoming a mark of a brand that has nothing to hide, and that is worth more than the smoothness you give up.

A practical checklist

You can stay ahead of this with a few habits built into production rather than bolted on after:

1. Know which assets are in scope. Realistic AI-generated or manipulated depictions of people and events are the sensitive category. A fully synthetic spokesperson is in scope; a stylised, obviously-illustrated graphic generally raises less concern.
2. Decide your disclosure language once. Agree a short, on-brand way of saying "this was made with AI" and apply it consistently, so it reads as a brand standard rather than a panic label.
3. Preserve provenance metadata. Keep the machine-readable markers your tools produce intact through your edit and publishing pipeline, rather than stripping them in finishing.
4. Make it part of the brief, not the cleanup. Decide disclosure at the briefing stage alongside the message, so the asset is compliant by design and you never ship something you then have to caption defensively.
5. Document your approach. A short internal note on how and why you disclose is cheap insurance and makes any future review a formality rather than a scramble.

The shift worth planning around

The deeper point is that transparency is becoming part of the product, not an obligation sitting next to it. The brands that adapt fastest will not be the ones with the best legal team; they will be the ones that decided clear, confident disclosure is simply how they want to sound. Done that way, compliance stops being a brake on AI production and becomes one more thing your audience can trust about you. That is the version of this worth building toward: not the minimum the law allows, but the posture that makes the law a non-event.

*This article is general information for marketing teams, not legal advice. Confirm how the EU AI Act and national rules apply to your specific case with qualified counsel.*

• European Union, Regulation (EU) 2024/1689 (the AI Act), Article 50 transparency obligations, 2024.
• European Commission, "AI Act: transparency obligations for generative AI," guidance, 2025.
• WARC, "Trust, transparency and synthetic media in advertising," 2025.