New rules for AI-generated content are now in effect in Denmark, while Norway is still working on implementing the AI Act. But does that mean all AI images and ads must be labeled? No, and the differences are important to understand.

‍

AI is already being used to create ad images, videos, text, voiceovers, and customer dialogue. From August 2, 2026 new transparency requirements in the EU AI Act will apply simultaneously.

For Danish companies, the rules are now in effect. In Norway, the situation is different. The government announced on August 4 that the Norwegian AI Act will be sent for a new consultation in the autumn of 2026, and the AI Act has therefore not yet been incorporated into Norwegian law.

However, that does not mean Norwegian companies can ignore the rules. The AI Act can also apply to businesses outside the EU if the output from the AI system is used within the EU. A Norwegian company creating content used in, for example, Denmark, may therefore fall under the regulations.

But first, an important clarification:

‍

The AI Act does not require all AI-generated marketing content to have a visible "made with AI" label.

‍

Must an AI-generated ad image be labeled?

‍

Not necessarily.

There are two different types of labeling requirements that are easily confused.

The first applies to the provider of the AI system. As a general rule, generative AI systems that create images, audio, video, and text, among other things, must be able to label AI-generated or manipulated content in a machine-readable format.

This can be achieved through metadata, watermarking, or other technical methods that make it possible to detect that the content is AI-generated.

This is primarily the responsibility of the AI system provider – it is not a requirement for the advertiser to write "AI-generated" in large text on the image. There are also exceptions, such as for standard editing where AI only performs an assistive function.

The second labeling requirement applies to the business that actually uses the AI content.

Deepfakes are particularly relevant here.

‍

Deepfakes must be clearly labeled

‍

If a company uses AI to create or manipulate images, video, or audio so that it significantly resembles a real person, object, place, or event – and could appear authentic – the content may be considered a deepfake.

In such cases, the audience must be clearly informed that the content is artificially generated or manipulated.

For marketing, the distinction is important.

‍

Imagine two ads:

Example 1:
You generate a clearly illustrative image of a futuristic store that does not exist.

This is not automatically a deepfake that requires visible labeling.

‍

Example 2:
You create a realistic AI video that makes an actual famous person appear to be endorsing your product.

Here, the risk of the content being perceived as authentic is much higher, making labeling requirements relevant.

The crucial factor is therefore not just whether AI was used, but what kind of content the AI has created and how the recipient might perceive it.

‍

What about AI-generated people in advertisements?

‍

This is a more interesting gray area.

A photorealistic AI-generated person is not necessarily a deepfake if the person does not represent an actual individual.

However, if the image or video gives the impression of documenting a real event, customer, employee, or situation that never actually took place, the company should be far more cautious.

The European Commission's guidance emphasizes, among other things, how realistic the content is, the message, the context, and what the audience can reasonably expect to be real.

This means that companies should not reduce their assessment to:

"Is the person real or AI?"

A better question is:

"Would an average customer perceive this as documentation of something that actually happened?"

‍

Must AI-generated ad copy be labeled?

‍

As a general rule, standard AI-assisted ad copy does not need a visible AI label just because ChatGPT, Claude, or another tool was used to write it.

The AI Act has a specific requirement for AI-generated or manipulated text published to inform the public about matters of public interest.

Even here, there is an important exception: if the text has undergone human review or editorial control, and an individual or organization holds editorial responsibility, the same labeling requirement does not apply.

For a standard product advertisement, newsletter, or landing page, this means you are not automatically required to include "written by AI."

However, if a business uses AI to mass-publish unedited information about politics, social issues, or other matters of public interest, the assessment changes.

‍

AI chatbots are also included

‍

For marketers, it is not just content production that is relevant.

The AI Act also requires that people be informed when they are communicating directly with an AI system, unless it is already obvious that they are speaking to an AI.

If your online store or company website uses an AI-based chatbot to answer questions, recommend products, or qualify leads, the customer should clearly understand that they are communicating with an AI, not a human.

This is particularly relevant now that more and more companies are integrating generative AI directly into their sales and customer service operations.

‍

Denmark and Norway have different statuses

‍

For Danish companies, Article 50 of the AI Act applies from August 2, 2026.

The Danish Agency for Digital Government has already published its own guidance on transparency requirements, including machine-readable labeling, deepfakes, and AI-generated text.

In Norway, the plan was previously for the AI Act to be implemented at roughly the same time as in the EU. This has changed.

On August 4, 2026, the government announced that changes to EU regulations mean the proposed Norwegian AI Act will be sent for a new public consultation in the autumn of 2026.

Norwegian companies should still start establishing routines now – especially if they market themselves in Denmark, Sweden, Finland, or other EU countries.

‍

Five things the marketing department should do now

‍

1. Map out where AI is used
Not just in text production, but also images, video, voiceovers, chatbots, and automated advertising tools.

2. Distinguish between AI assistance and synthetic content
Using AI to remove the background from a product image is different from generating a realistic event that never took place.

3. Be extra careful with real people
The use of AI to imitate customers, employees, celebrities, or other actual people should have a significantly higher threshold.

4. Maintain human editorial responsibility
AI can produce drafts, but the business should still verify claims, images, and messaging before publication.

5. Document how the content is created
Have a simple internal routine for which tools are used and what checks have been performed – especially for more realistic AI-generated content.

‍

AI in marketing does not require a warning on everything

‍

The new rules do not mean that companies have to stop using generative AI in their marketing.

Nor do they mean that every AI-generated image needs a large label.

Transparency is key where AI poses a real risk of misleading the public about what is authentic.

For Danish companies, these requirements are already relevant. For Norwegian companies, the Norwegian AI Act is still under development, but businesses operating within the EU should not wait for it to come into force in Norway before reviewing their procedures.

‍

Does your company use AI in its marketing?

‍

WeAssist helps Norwegian and Danish companies integrate AI into their marketing, content production, and digital workflows.

We can help establish effective workflows for AI-generated content, human quality assurance, and cross-market publishing.

Get in touch with WeAssist if you want to use AI more effectively – without losing control over quality, brand, or transparency.

‍