When is it mandatory to disclose the use of AI in advertising?

2 August 2026 marks the beginning of a new phase in the application of the European regulation on artificial intelligence (AI Act): the transparency obligations become applicable. Anyone using an AI system will also have to clearly indicate this.

As a reminder, the AI Act entered into force on 1 August 2024 and is being applied progressively.

On 20 July 2026, the European AI Office finally published the final version of its “Guidelines on transparency obligations for providers and deployers of AI systems”. These guidelines are intended to explain and harmonise the application of Article 50 of the AI Act. They do not create any new obligations, but clarify how the European Commission interprets the transparency obligations laid down in the regulation in order to ensure their uniform application across all Member States.

These transparency obligations are intended to make it easier for the public to recognise when they are interacting with an AI system or when content has been generated or manipulated by artificial intelligence.

For advertisers, agencies and other communication professionals, the most important obligation concerns content that may be classified as a “deepfake”.

When is content considered a deepfake?

The AI Act defines a “deepfake” as AI-generated or manipulated visual, audio or video content that resembles an existing person, object, place, entity or event and could be wrongly perceived as authentic or truthful.

According to the Commission’s guidelines, the following elements must be assessed together:

• the content displays a sufficient degree of resemblance to the subject represented;
• the content must depict or resemble a person, object, place, entity or event that exists, or that could plausibly exist or have existed in reality;
• the content must be likely to be wrongly perceived as authentic or truthful.

The assessment must take into account, in particular, the degree of resemblance, the message conveyed, the context in which the content is distributed, the target audience and its expectations.

The concept of a deepfake is therefore not limited to imitating a celebrity or an existing person. A completely fictional person presented as a real consumer, expert or influencer may fall within its scope when their portrayal is sufficiently realistic and the public could believe that the person or experience is authentic.

The same may apply to an artificial product demonstration or a realistic depiction of a result that never actually occurred.

Depending on the circumstances, a disclosure may be required for:

• a realistic virtual influencer claiming to personally use a product;
• a synthetic voice imitating a real person;
• a fictional person giving a testimonial presented as authentic;
• a realistic demonstration generated entirely by AI and presented as real;
• an image showing a result or performance that was not achieved in reality.

These situations must always be assessed on a case-by-case basis.

The mere fact that content has been created or manipulated using AI is not sufficient for it to be automatically considered a deepfake. Content that is clearly imaginary, caricatural or unrealistic will generally not be considered a deepfake if the public cannot reasonably perceive it as an authentic or truthful representation.

Not all AI retouching constitutes a deepfake

The mere use of an AI function is not sufficient for content to be classified as a deepfake.

Limited technical modifications, such as reducing background noise, correcting colours, cropping or using AI-generated backgrounds, are not subject to the transparency obligations when they are not likely to mislead the public as to the content’s authenticity or truthfulness.

However, when AI alters the public’s perception by giving the impression, for example, that a product is larger, performs better or is more attractive than it actually is, the content may fall within the rules applicable to deepfakes, without prejudice to the provisions relating to misleading advertising.

How should the public be informed?

When advertising content constitutes a deepfake, the fact that it has been artificially generated or manipulated must be disclosed no later than when the public is first exposed to the content.

The disclosure must be clear, perceptible, distinct and understandable. It must also comply with the applicable accessibility requirements.

The public must not be required to consult metadata, use a technical tool, click on a menu or carry out a specific search to discover that the content has been generated or manipulated using AI.

The AI Act does not impose a single form of wording. Depending on the format and nature of the creation, the following disclosures could, for example, be used:

“This image was generated by artificial intelligence.”

“This video contains images generated or manipulated using AI.”

“This person is virtual and was created using AI.”

“This voice was generated or imitated using AI.”

“This content was artificially manipulated using AI.”

Vague wording such as “digitally created”, “made using technology” or “enhanced content” may not indicate the use of artificial intelligence clearly enough.

For an image, the disclosure may be incorporated into the visual or placed directly next to it. For a video, it may be clearly displayed from the beginning. For audio content, an audible announcement may be necessary.

The size, contrast, display duration and placement must be appropriate for the medium and the audience. Particular attention is required when advertising is aimed at children, vulnerable people or audiences with limited familiarity with AI technologies.

However, these transparency obligations do not exempt advertisers from complying with all other rules applicable to advertising. The content, message and presentation of the advertisement must remain fair, truthful and non-misleading.

Next steps in Belgium

The Advertising Council, together with its partners, is currently preparing additional practical guidelines based on the final guidelines published by the European Commission on 20 July 2026. Their purpose is to help advertisers, agencies, media organisations, platforms and production partners apply these new obligations in the Belgian advertising context.

These guidelines will be published in the coming weeks.

They will complement the twelve principles for the ethical use of AI in advertising, which have already been developed with the main organisations in the Belgian advertising industry.

Find out more

Consult the Guidelines on transparency obligations for providers and deployers of AI systems published by the European Commission.