ASA warns on AI-generated content and deepfakes

The Advertising Standards Authority (ASA) has published new guidance on the use of AI-generated content and deepfakes in advertising, reinforcing that existing rules apply in full, regardless of how ads are created.

As AI tools become increasingly embedded in marketing processes, the ASA is signaling a robust, technology-neutral approach to enforcement.  It makes the following key points:

The CAP Code applies to AI-generated ads

The CAP Code is media-neutral. Ads created using AI are subject to the same rules on misleadingness, offence and social responsibility as any other marketing content. The ASA demonstrated in its ruling in Polyverse Inc that harmful, offensive or socially irresponsible AI‑generated images can breach the CAP Code like any other image would. A mobile ad promoting the AI photo editor AI Mirror appeared in September 2023 within the game Pocket Champs. The ad showed a woman depicted in an anime-influenced style, posed in a bent-forward position wearing revealing shorts pulled up to expose her buttocks. On-screen text encouraged viewers to try the app's AI functionality. The ad was deemed socially irresponsible and likely to cause serious offence, with the ASA also concluding that its placement in a gaming environment was inappropriate regardless of the specific game.

Deepfakes create heightened risk

AI-generated depictions of real individuals, particularly celebrities, must not mislead consumers into believing there is a genuine endorsement. Use of a person’s likeness may also raise separate legal risks (for example, IP and passing off), even if outside the ASA’s remit.

Advertisers remain responsible for AI bias

If AI tools produce harmful or stereotypical content, liability sits with the advertiser. Outputs should be reviewed to ensure they are socially responsible, non-discriminatory and appropriate for the audience and context. This was illustrated by the Cosmos ruling, where the ASA told the advertiser to make sure that future ads were socially responsible and did not cause serious offence, including by featuring a harmful gender stereotype that objectified women. An ad for an AI video generation app appeared within a mobile game in April 2025. It showed an image of a man labelled "me" placed beside an image of singer Sabrina Carpenter labelled "My crush". An AI-generated video then depicted the two kissing, accompanied by a prompt to "kiss your celebrity crush". A voiceover encouraged users to upload photos so the app could generate a kissing video with a celebrity of their choice within seconds. The ASA found that the invitation to generate intimate deepfake videos using images of real women to be both harmful and offensive, reinforcing the impression that it was acceptable to objectify women through technology. 

Automation does not shift accountability

Using automated ad creation or placement tools does not transfer responsibility. Advertisers remain accountable for both content and targeting decisions. The ruling in Stripe & Stare was a prime example of this.  A display ad for Stripe & Stare's pyjama set appeared on the gaming website IGN in December 2022. It showed a young female model in an open pyjama top that left her chest exposed, posing with her arms raised and hands touching her face. The ASA found that the ad was socially irresponsible and likely to cause serious offence.  Importantly, it stated that even though the ad was produced using Google’s automated tools, the advertiser was still primarily responsible for ensuring compliance with the CAP Code. 

What does this mean for advertisers?

The collective message from these rulings is unambiguous. AI offers significant creative potential, but advertisers bear ultimate responsibility for every piece of content that reaches the public, however it was generated. Delegating production to an algorithm does not delegate accountability. If a synthetic celebrity likeness is convincing enough to mislead consumers, the regulatory consequences are the same as for any other misleading advertisement. 
In short, advertisers would be well advised to review AI-generated content with the same rigour they would apply to any other creative output — preferably before it goes live.