AI is already baked into the way marketing teams work. It helps resize creative, clean up audio, draft copy, generate images, translate content and turn one campaign idea into a dozen different assets.
From 2 August 2026, some of that work will come with new transparency requirements under the EU AI Act.
The good news is that you don’t need labels on every piece of AI-assisted content. The Act simply asks marketers to be clearer when content could reasonably be mistaken for something real.
What is the EU AI Act?
The EU AI Act is a wide-ranging law designed to regulate how artificial intelligence is developed and used.
It follows a risk-based approach. The greater the potential harm, the stricter the requirements.
Most ordinary marketing activity does not fall into the Act’s highest-risk categories. The section marketers are most likely to encounter is Article 50, which introduces transparency obligations for certain AI-generated and manipulated content from August 2026.
This includes:
- Deepfakes and other realistic synthetic media
- Certain AI-generated public-interest content
- Customer-facing AI systems where it is not obvious that someone is interacting with AI
- Emotion-recognition and biometric-categorisation systems
The important distinction is that these rules do not apply equally to everything created with AI.
For creative teams, the biggest concern is content that depicts a person, place, object or event in a realistic way, despite that content being generated or materially altered by AI.
Think of a CEO appearing to say something they never recorded, a fabricated photograph of a real event or an AI-generated product demonstration presented as authentic. That is very different from using AI to tighten a headline, resize an image or create an obviously illustrated campaign visual.
The law is not asking, “Did AI touch this?”
It is closer to asking, “Could someone reasonably mistake this for reality?”
How does this impact marketers?
This should be on your radar if your team:
- Creates photorealistic AI images or videos
- Uses synthetic presenters, face swaps or cloned voices
- Alters footage of real executives, employees or customers
- Publishes AI-generated information on public-interest topics
Location alone is not a reliable way to decide whether the rules matter. The rules may apply even when the company or agency producing the work is based outside Europe. For global brands, a campaign produced in the US but published to European audiences may still fall within scope.
This is particularly relevant for global social channels. A post may be created by one central team but viewed, shared and republished across multiple markets.
What actually needs to be labelled?
There will inevitably be grey areas, but the following examples offer a useful starting point.
Label it
- A realistic AI video of a real executive
- A cloned voice used as though it belongs to a real spokesperson
- An AI-generated photograph of an event that never happened
- A materially altered image that changes what appears to have occurred
- Public-interest information generated by AI without meaningful human editorial review
Probably not
- An obviously fantastical or illustrated image
- Basic cropping, resizing or colour correction
- Removing a small background distraction without changing the meaning
- Copy drafted with AI but substantively reviewed, fact-checked and approved by a person
- AI used internally for research, ideation or early concepts
Context matters too.
A surreal visual used in a playful campaign is unlikely to be mistaken for documentary evidence. The same visual presented as a real customer story or news event carries a different risk.
Where disclosure is required, it should be clear when the audience first encounters the content. It should not be hidden at the bottom of a long caption or buried on another webpage.
The EU has also developed optional labels including:
(Source: Screenshot taken from the European Commission)
Brands do not necessarily have to use those exact icons, but they provide a useful model for creating a consistent disclosure system.
What should marketing teams do now?
There is no need to relabel your entire content library or put a compliance review behind every asset.
But teams should start putting a few basics in place.
1. Audit where AI is used
Include image generation, voice cloning, synthetic presenters, face swaps, chatbots and AI-assisted public-facing copy.
2. Create a disclosure standard
Decide what wording or icon the brand will use and where it should appear across social, websites, advertising, video and audio.
3. Add AI to the approval process
Record which tools were used, what was generated or altered and who reviewed the final asset.
4. Clarify who owns the decision
Agree whether the brand, agency, production partner or platform is responsible for identifying and adding the disclosure.
5. Use judgement
When an asset sits close to the line between creative interpretation and apparent reality, transparency is usually the safer choice.
A new creative consideration
AI is evolving quickly, and regulation will continue to evolve alongside it. The EU AI Act is unlikely to be the last piece of AI regulation marketers will navigate.
Ultimately, this isn’t just about compliance. It’s about setting expectations for how your brand uses AI.
Rather than treating transparency as a box-ticking exercise, marketers should see it as an opportunity to build trust while continuing to experiment with the technology. The goal isn’t to create less with AI – it is to use it more responsibly.
