Using AI imagery and voice clones in ads: UK rules
The UK has no separate rulebook for AI ads. The CAP Code applies as it always did, and for voices and faces the protection comes from contracts and consent. This is what that means for a brief.
In short: There are no AI-specific UK advertising rules. The CAP Code applies however an ad is made, and the advertiser is responsible for the output. There is no blanket duty to label AI use, but a misleading image is still misleading. For a voice or face, UK law has no personality right, so written consent does the work.
UK advertising has no separate rulebook for AI. The CAP Code applies to an ad whether a camera, an illustrator or a model made it, and the advertiser answers for the result. Voices and faces are a different problem, because UK law gives people no general right over their own voice or likeness. This page is general guidance, not legal advice. For a campaign with real money behind it, a solicitor who knows advertising law should read the brief.
What the ASA and CAP have actually said
The ASA published four points for advertisers on 11 June 2026, and law firms summarised them the following week. They are worth knowing as the regulator's own position:
- The Code is media-neutral. The ASA's Polyverse ruling shows that an AI-generated image can breach the Code if it is harmful, offensive or socially irresponsible.
- Deepfakes are not exempt. An AI-made celebrity endorsement that looks real enough to mislead people is likely to breach the Code. The ASA does not regulate personality rights or intellectual property, and says advertisers should still consider those risks and take legal advice before using a public figure's likeness.
- Bias is a compliance issue. In the Cosmos ruling, in-game ads for an AI simulator app that objectified women breached the Code. Advertisers remain responsible for checking AI output is socially responsible, free from harmful stereotypes and suitable for the audience and placement.
- Automation does not move responsibility. In the Stripe & Stare ruling, the ad was produced by Google's automated tools and the ASA still held the advertiser primarily responsible, including for the placement.
The lesson across all four is the same: the tool is not a defence. Treat AI output as you would a freelancer's first draft, and review it before it goes out.
Do you have to say it is AI?
Not as a general rule. CAP's guidance, summarised by RPC and Osborne Clarke in 2025, says the UK has no blanket legal duty to disclose AI use in ads. CAP has said it could change its approach if that became proportionate. It suggests two questions:
- Would people be misled if you did not say so?
- Would a disclosure clarify the message, or contradict it?
CAP also says that using AI for edits comparable to traditional post-production does not by itself call for disclosure, provided nothing is materially misleading. A disclosure can help where the AI element changes how people read the ad, such as a comedic deepfake or an influencer who is openly artificial. It cannot rescue a message that misleads. CAP's example is an AI-generated cosmetic result that does not reflect real-world outcomes: disclosure alone is very unlikely to cure that.
Industry principles can set a higher bar than the law. The ISBA and IPA principles from 2023 ask for transparency where AI plays a significant role that consumers may not notice. If you work for a big brand or an agency with its own policy, follow that policy.
AI imagery: the three things that go wrong
An image that overstates the product. A generated hero shot of a skincare result, a kitchen or a view can imply a result the product does not give. Compare each frame with what the product does on a normal day.
A fake customer. The ASA's testimonial rules say you must hold documentary evidence that a testimonial is genuine, along with contact details for the person giving it, and must not use one without permission. A synthetic person presented as a real customer cannot meet that test. An invented customer quote is a misleading claim, however good the image looks.
An ad nobody reviewed. Automated campaign tools generate versions at speed. Someone with authority has to approve what they produce, including the placement.
Voice clones and likeness
For a voice or a face, UK protection is thin. Pinsent Masons, in an analysis published on 15 September 2026, says no single UK law covers voice and likeness cloning. The routes are passing off, which needs established goodwill, misrepresentation and damage; copyright, which protects a specific recording and not a voice; and trade marks, which do not cover the qualities of a person's voice. Reed Smith, writing on 30 September 2026, says English law has no standalone right to control the commercial use of a name, image, likeness or voice.
In March 2026 the government said it would explore options, including a possible new personality or digital replica right. Reed Smith reports that a consultation was promised for summer 2026, had not been published by 30 September 2026 and had no revised date. Check the position again before you commit to a campaign, because it may have moved.
Data protection is the second route. Pinsent Masons notes that recording a person's voice or appearance and feeding it to an AI model will often count as processing personal data. Consent has to be specific to the purpose and can be withdrawn. A replica can only be used for the purpose set at the start.
In practice, a signed agreement does the work the law does not. The performer side is covered in what performers and the people who hire them can do now. A usable agreement answers these questions in writing:
- What recordings are made, who keeps them and for how long.
- What the voice or likeness may be used for: which ads, media, territory and period.
- Whether it may be used to train or build a synthetic voice. If not, say no in writing. If yes, say so and agree separate payment for it.
- What happens at the end of the term: deletion, by when, and who confirms it.
- How the person can withdraw consent, and what happens to ads already live.
- Whether the ad will say the voice is synthetic.
Our voice and likeness checklist sets the same points out as a form to complete.
What to put in a supplier brief
If an agency, studio or freelancer is making the ad, the brief is where responsibility gets settled. Four lines cover most of it:
- Say whether AI may be used, and for what. "AI-assisted editing and clean-up is fine; generated people are not" is a clear instruction. Silence leaves it to the supplier's habits.
- Ask for a list of every AI tool used on the job, so you know whose terms apply to the output and where recordings were processed.
- Ask who owns the output and what the tool's terms allow. Some tools restrict commercial use or keep rights in what they produce.
- Ask for consent paperwork for any real voice or face, signed before the work starts, and keep a copy yourself.
A supplier who can answer those quickly has done this before. One who cannot may be learning on your campaign.
A short table for the brief
| Task in the ad | Main risk | What to do |
|---|---|---|
| Generated background or abstract image | Low, if it does not change what the product does | Review it like any other creative |
| Generated product result | Misleading claim | Compare with real performance, or shoot it |
| AI avatar or synthetic customer | Fake testimonial, misleading impression | Do not present as a real customer |
| Cloned voice of a performer | Consent, data protection, passing off | Written consent, scope, deletion date |
| Lookalike of a public figure | Misleading endorsement, passing off | Do not use without the person's agreement |
| Automated ad variants | Advertiser still responsible | Human approval before launch |
Who signs it off
Name one person who approves AI-assisted creative, and keep a short record of what was generated, with which tool, and who checked it. If an ad is challenged, that record is what you show. Granary's AI Studio includes governance you can show a board among its capabilities, if you want that process set up as part of a wider AI plan. If the question is whether AI is the right tool at all, when to use AI and when to shoot it for real is the better starting point.
Sources
- ASA and CAP, AI and deepfakes: four things advertisers need to know before they hit run, read 5 November 2026
- Lewis Silkin, ASA warns on AI-generated content and deepfakes, read 5 November 2026
- RPC, CAP issues guidance on use of AI in ads, read 5 November 2026
- Osborne Clarke, AI in advertising: revolutionising creativity and compliance, read 5 November 2026
- Pinsent Masons, Digital replicas and the personality rights debate, read 5 November 2026
- Pinsent Masons, Digital replicas: the data protection questions for businesses, read 5 November 2026
- Reed Smith, AI and rights of publicity in the UK, read 5 November 2026
- ASA, Testimonials and endorsements, read 5 November 2026
Related guides
- Ofcom's streaming access code: what 80% subtitles, 10% audio description and 5% signing mean for producers
- The less healthy food ad rules, nine months in: what the rulings mean for your next shoot
- Music licensing for brand video in the UK
- Releases: talent, location, property, minors
- All guides and tools
The picture at the top of this page is an illustration, not a photograph of a client or a real job.
Questions we get asked
Do UK ads have to say they were made with AI?
There is no blanket duty. CAP's guidance is to ask whether the audience would be misled without a disclosure, and whether the disclosure clarifies the message. A label does not cure an ad that misleads.
Can I clone a voice for an ad in the UK?
There is no UK law that gives a person a general right over their voice, so you rely on a signed agreement that says what the voice can be used for, where and for how long. Without consent, you risk passing off, data protection and misleading-endorsement problems.
Who is responsible if an AI tool makes a non-compliant ad?
The advertiser. In a ruling on an ad made with Google's automated tools, the ASA held the advertiser primarily responsible for compliance with the CAP Code, including the content and where it appeared.
Is it legal to use an AI-generated celebrity in an ad?
The ASA says an AI depiction that looks real enough to mislead people into thinking a celebrity endorses the product is likely to breach the CAP Code. Separate legal risks, such as passing off, sit outside the ASA's remit and need legal advice.
granary.digital · 01243 884993 · studio@granary.digital · The Old Grainstore, Adsdean, Chichester, West Sussex, PO18 9DN · Prices exclude VAT unless marked inc VAT