What do companies need to know about disclosing use of AI in marketing in light of the changes introduced by the EU AI Act, and how does this differ for the UK? We have set out below a summary of AI transparency obligations for marketing in the UK and EU.
- UK position
There is no specific requirement to label content as AI-generated. However, consumer protection and advertising rules will apply. For example, the Advertising Standards Authority has published guidance such as AI and Deepfakes: Four Things Advertisers Need to Know Before They Hit “Run” and Disclosure of AI in Advertising: Striking the Balance Between Creativity and Responsibility.
The key question is whether the use of AI could mislead consumers. If the AI element would affect a consumer’s understanding of the product or influence their decision-making, then the fact that the content is AI-generated may be considered “material information” and should be disclosed. If not, there is typically no obligation to label it.
In practice, this means that background use of AI in an ad (for example, generating a background image that does not relate to product performance or claims) will usually not require disclosure. However, if AI is used to create or enhance elements that contribute to product claims or consumer expectations (e.g. product results, effects etc.), then the risk of a misleading omission increases and disclosure should be considered.
- EU position
In the EU, consumer protection and advertising rules similarly apply. However, the EU now has a specific regime under the AI Act. We have summarised the new rules as follows:
- AI-generated marketing content is not automatically subject to disclosure
A common misconception is that all AI-generated marketing content will need to be labelled. The position is more nuanced.
The transparency obligation most relevant to marketing is contained in Article 50(4) AI Act, which applies where an AI system generates or manipulates image, audio or video content constituting a “deep fake”. The question is therefore not simply whether AI has been used, but whether the resulting content falls within the AI Act’s definition of a deep fake.
The Commission’s Guidelines on transparency obligations for providers and deployers of AI systems indicate that a deep fake will generally involve three cumulative elements:
- An appreciable resemblance or realism;
- A depiction of persons, objects, places, entities or events that exist, could exist, or plausibly could have existed; and
- Content that may falsely appear authentic or truthful to the intended and reasonably foreseeable audience.
Importantly, no intention to deceive is required. The assessment is based on how the relevant audience may perceive the content. This lowers the threshold in practice (particularly where audiences may be more vulnerable or less able to detect synthetic content).
- Realistic synthetic people are a key area of focus
One important clarification from the Guidelines is that the content does not need to depict a real person.
The Commission expressly indicates that realistic AI-generated persons, avatars, influencers, executives and campaign models can fall within scope where they resemble someone who could plausibly exist and appear authentic to consumers. Examples provided by the Commission include:
- AI-generated celebrity influencers;
- realistic synthetic presenters;
- AI-generated company CEOs; and
- realistic synthetic avatars.
As a result, brands making use of realistic AI-generated humans in advertising, influencer campaigns or social media content should carefully assess whether disclosure obligations may arise.
- Minor or technical AI edits are generally lower risk
The Guidelines distinguish between substantive AI manipulation and routine production or editing activities.
Examples of modifications that will generally not trigger deep fake concerns include:
- colour correction;
- lighting adjustments;
- noise reduction;
- technical enhancements;
- image compression; and
- other minor cosmetic edits that do not materially affect how the content is perceived.
Similarly, the Guidelines suggest that AI-generated background elements or environmental imagery are less likely to fall within the deep fake concept where they do not materially affect consumer understanding of the product or message being communicated.
- Product imagery requires particular attention
The Commission specifically highlights AI-generated product imagery as an area where disclosure obligations may be more likely to arise.
The Guidelines explain that AI-generated depictions of products may constitute deep fakes where the content affects consumer perception regarding:
- the appearance of the product;
- the characteristics of the product;
- the quality of the product; or
- how the product functions or is used.
This is particularly relevant for sectors such as beauty, fashion, luxury goods, consumer products and food and beverage, where marketing often relies on visual representations of products and outcomes.
- If content is a deep fake, clear disclosure is required
Where content falls within the deep fake definition, deployers of the AI system (i.e. the organisation using the AI-generated content in a commercial context) must disclose that the content has been artificially generated or manipulated.
The accompanying Code of Practice on Transparency of AI-generated Content proposes a standardised approach, including:
- “AI Generated” labels;
- “AI Modified” labels;
- an EU AI icon; and
- disclosures displayed in a clear and distinguishable manner at the point of first exposure.
While compliance with the Code is voluntary, the Commission has indicated that adherence is expected to provide a recognised route to demonstrating compliance with Article 50.
- Marketing content is not necessarily exempt because it is creative
The AI Act provides a lighter disclosure regime for content that forms part of an evidently artistic, creative, satirical, fictional or analogous work.
However, the Guidelines make clear that commercial advertising will not automatically benefit from this exception. The Commission specifically identifies the following as examples that may still require disclosure:
- realistic synthetic influencers;
- AI-generated advertising personalities;
- AI-generated product demonstrations; and
- AI-generated commercial content focused on showcasing product functionality.
In practice, many brand marketing campaigns are likely to be assessed under the ordinary deep fake rules rather than the reduced disclosure regime for creative works.
- Consumer protection considerations remain important
Even where Article 50 does not strictly require disclosure, businesses should continue to assess whether the use of AI-generated content could be misleading under consumer protection or advertising laws.
In particular, regulators may be concerned where AI-generated content influences consumer perceptions regarding:
- product performance;
- efficacy;
- quality;
- results; or
- characteristics that are material to purchasing decisions.
As a result, disclosure may sometimes be advisable from a consumer protection perspective even where a strict AI Act obligation is uncertain.
- Timing of EU AI Act transparency obligations
These transparency rules apply from 2 August 2026.
There is a grace period for providers of generative AI systems placed on the market before 2 August 2026 to bring their systems into conformity regarding the marking and detection obligation by 2 December 2026.
For deepfakes generated before 2 August 2026, there is no mandatory retroactive labelling obligation but the Commission encourages such labelling.
Don’t hesitate to reach out to any of us to discuss more on what this means for your organisation.
