Not long ago, the question of AI in advertising mostly came down to how much faster and cheaper an ad creative could be made. Today businesses face a different question: where does ordinary use of the technology end and the legal duty to tell consumers that they are looking at the work of artificial intelligence begin?
This matters most for international campaigns. The same AI-generated ad can run in the UK, the EU and other jurisdictions at once, while transparency requirements already differ between them.
The main mistake is to think that a small “AI-generated” label settles the matter. In many cases the problem is not that the ad was created by AI, but what the ad tells the consumer.
AI does not exempt advertising from the usual rules
The UK has no general rule requiring every image, video or text created with AI to be specifically labelled. That does not mean an advertiser can use AI without legal risk.
The position of the UK regulator, the ASA/CAP, rests on media neutrality: the way an ad is made does not change the rules that apply to it. AI may change how ads are produced, but it does not change the requirements for accuracy and compliance with the advertising rules.
So a UK advertiser should ask not “do I need to label AI?” but “could a consumer draw the wrong conclusion from the way I have used AI?”. That is where the main legal risk lies.
The riskiest case: AI creates something that does not exist
Consider a few examples:
- a company photographs a real cosmetic product but uses generative AI to alter the model’s skin so the result looks far more dramatic than it could really be;
- a car manufacturer uses AI to create an image of a model in a specification that does not exist;
- an online shop shows a product in a colour or design in which it was never sold.
In all these cases, a note saying “image created with AI” does not solve the problem. If an ad image misleads consumers, it stays misleading whether it was made by a designer, a photographer or a neural network.
The ASA is explicit on this: AI is not an exception to the general rules on misleading advertising, and using the technology does not excuse false claims or misleading images.
This is a key point for business: transparency about AI and the accuracy of the ad are two separate legal questions. You can honestly say an image was made by AI and still breach the rules on misleading advertising.
What if AI simply “cleaned up” a photo?
The situation is quite different here. AI can remove an unwanted object from the background or adjust lighting, colour, weather and other minor details. That use does not necessarily call for separate disclosure.
The ASA draws a parallel with traditional post-production: if a change is not material to how the ad is understood, the consumer gets no new information that affects their decision. For example, if a bin was accidentally left in the background of an athlete’s photo and AI removed it, this is unlikely to create a separate duty to disclose the use of AI.
But if AI removes not a bin but wrinkles, skin blemishes, excess weight or other features of a person or product that matter to the message of the ad, the assessment changes.
The line is not between “AI was used” and “AI was not used”. It is between a material and a non-material change to the advertising message.
In the EU, the picture changed on 2 August 2026
For companies working with European audiences, the question has become even more practical. The transparency requirements of Article 50 of the EU AI Act began to apply on 2 August 2026. In July 2026 the European Commission published separate guidance on how they apply in practice.
One key provision covers AI-generated or AI-manipulated content that could be taken as real, including deepfakes. EU law uses a broader concept than the usual idea of a deepfake as a fake video of a celebrity. A problem can arise even where the image shows no real person at all.
For example, a company could use AI to create:
- a photograph of a hotel that does not exist;
- an image of a product that has never been photographed;
- a realistic picture of a location that does not exist;
- a video featuring a fictional person;
- a visualisation of an event that never took place.
If such material looks real or gives an impression of factual accuracy, the transparency rules cannot be set aside just because “the person is made up”.
That is why the European Commission stresses the risks of deception, manipulation and impersonation that come from the growing difficulty of telling AI-generated content from authentic content.
Is every AI image of a product a deepfake?
No, and this is one of the hardest practical boundaries. If AI has only lightly edited a photo, that does not automatically make it a deepfake.
But if a product image is changed so much that it suggests a different product, one that is better, more functional or has features it does not have, the risk rises sharply.
This is where an advertising creative stops being a design question and becomes a compliance question. A lawyer has to assess not the fact that AI was used but the final result and the impression it gives the average consumer.
A UK company can also face EU requirements
For international business this is probably the most important practical point. A company being based in the UK does not mean its campaign is automatically outside European regulation.
If AI content is available to an audience in the EU, you need to assess separately whether the relevant AI Act provisions apply. For global campaigns, the place where a company is registered is not always enough to determine the regulatory requirements. In practice, one campaign may need a different compliance approach depending on where it is shown.
And it is not only about images
Another underrated issue is AI-generated text. Article 50 of the AI Act covers more than visual content. The European Commission specifically points to AI-generated text published to inform the public on matters of public interest, where the text has not been through human or editorial review.
For advertisers this can matter, for example, when AI produces content on health, consumer safety, the environment or other topics of public interest. So the line “it’s only text, not a deepfake” does not always close the question.
Labelling must be real, not just formal
Even where disclosure is needed, it is not enough to put it somewhere at the bottom of the page. The European Commission stresses that information about AI content must be made available to people properly and in good time. This is especially important for video and social media, where a user may start watching after the first second.
Advertisers should therefore think not only about the wording of the label, but also about how visible it is, when it appears and how a specific audience will notice it. A formal disclosure that nobody sees may fail to do the job the requirement was meant to do.
What changes for advertisers
AI moves part of the legal control from the stage of “can we use this tool?” to the stage of “what exactly did we publish with it?”. A company can:
- use a perfectly legal AI tool and produce unlawful advertising with it;
- have no duty to label a particular AI edit and still be responsible for a misleading result;
- honestly add an AI label and still breach the rules if the product shown does not match the real one.
Conversely, AI can be used in a way that creates no significant risk for consumers at all.
For that reason, a short review of each creative is more practical than a universal “label / don’t label” checklist. Five questions are worth answering:
- What exactly was created or changed with AI?
- Does the change affect how the product is perceived?
- Could a consumer take the artificial for the real?
- Which audience will see the ad?
- Does the relevant jurisdiction require disclosure?
Conclusion
Artificial intelligence has not created a separate world of advertising rules. It has made an old problem, the accuracy of the advertising message, considerably harder.
In the UK, the focus remains on whether an ad is misleading and whether it complies with the general advertising rules. The ASA confirms that AI-generated advertising gets no separate exemption from them. In the EU, since 2 August 2026, special AI Act transparency requirements apply on top for certain categories of AI content, including deepfakes and some AI-generated publications.
The main takeaway is simple: the legal risk comes not from the fact that AI was used, but from how its output changes the advertising message and how consumers perceive it. AI advertising therefore needs legal review as well as creative control, especially when a campaign runs in several jurisdictions at once.
Need help with the use of AI in advertising?
If your company uses AI to produce advertising materials, it is important to assess not only what the tool can do technically but also the legal consequences of using it: whether the content could mislead consumers, whether the use of AI must be disclosed, what requirements apply in the countries where the ad is shown, and whether the labelling meets them.
Our team helps businesses assess these risks and review advertising materials against the requirements of the relevant jurisdiction. If you already use AI in marketing or are planning an international campaign, get in touch. We will help you work out which requirements apply to your case and set up an advertising process without unnecessary legal risk.