ASA Rules on AI-Generated Advertising

If you've already decided AI tools are a useful way to produce ad images and copy faster, the next question is whether the ad itself still holds up under the same rules that applied before AI was involved at all. It does, and that's the part that catches businesses out.

Why "AI made it" changes nothing about the rules

A marketing manager at a 15-person beauty brand using AI to generate product photography and ad copy is a common real scenario. If that AI-generated image shows a "before and after" skin result that exaggerates what the product actually delivers, the ASA treats that claim exactly the same as if a photographer and copywriter had produced the same exaggerated result by hand. The advertising codes are written around the effect an ad has on a consumer, not the tool used to produce it.

This means the core advertising principles still apply in full: an ad must not mislead by inaccuracy, ambiguity, exaggeration, or omission, and any claim (visual or written) must be capable of being substantiated. An AI-generated image that implies a result the product cannot reliably deliver is a substantiation problem, regardless of how the image was produced.

When you need to disclose that AI was used

There's no blanket UK legal requirement to disclose AI involvement in every ad. CAP's guiding principle instead asks a practical question: would the audience be misled if the use of AI isn't disclosed? Where an AI-generated element features prominently in an ad and its involvement isn't obvious to a reasonable consumer, disclosure is expected to avoid misleading them about what they're actually looking at.

A stylised AI-generated background image behind a genuine product photo is a very different case to an AI-generated "customer result" image presented as if it were a real photograph of a real outcome. The second case is much more likely to need disclosure, or to need rethinking entirely if the underlying claim can't be substantiated regardless of disclosure.

Deepfakes and AI-generated testimonials

The ASA has specifically warned advertisers over AI-generated content and deepfakes, an area where the risk is highest: an AI-generated "customer" giving a testimonial, or an AI-generated version of a real person's likeness used in an ad without genuine consent, raises both an advertising standards problem and a separate legal risk around consent and image rights. Neither of those risks disappears because the content was AI-generated rather than filmed.

What enforcement actually looks like

The ASA has signalled that both AI-generated advertising and advertising that references AI capabilities are on its radar for 2026, with further CAP guidance or enforcement action reasonably expected during the year. The ASA is also scaling up its own AI-based Active Ad Monitoring system specifically to proactively identify non-compliant ads at scale, rather than relying solely on consumer complaints. A business assuming AI-generated content is less likely to be scrutinised has that backwards: it's an active monitoring focus, not a blind spot.

Where this trips up smaller marketing teams specifically

A common pattern on a small marketing team is treating AI image generation the same way as stock photography: pick something that looks good, drop it in, move on. Stock photography rarely makes a specific factual claim about a product's effect. AI-generated "result" imagery often does, even when nobody intended it to. The risk isn't the tool, it's skipping the same substantiation check a team would normally apply to a claim written in words.

Building a short internal check into the existing sign-off process, the same one already used for any other ad claim, is usually enough. The mistake to avoid is treating AI-generated content as a separate, lower-scrutiny category simply because no photographer or copywriter physically produced it.

What you can do about it

Three practical steps: treat every AI-generated image or claim exactly as you would a human-produced one before it goes live, ask whether the result shown could actually be substantiated, and disclose AI involvement where a reasonable consumer would otherwise be misled about what they're looking at, particularly for testimonials, before-and-after results, or any depiction of a real person.

This is general guidance, not a compliance clearance for any specific ad. Confirm your position with CAP's published advertising guidance or a qualified adviser before running an ad campaign that leans heavily on AI-generated content.

FAQ

Related reading: our AI governance by region.

Do I always have to label an ad as AI-generated?

Not always. There's no blanket legal requirement to disclose AI use in every ad. Disclosure matters where a reasonable consumer would be misled about what they're seeing without knowing AI was involved, particularly for testimonials or result claims.

Can I use an AI-generated image of a person in an ad?

This carries real risk beyond advertising standards, particularly if the image resembles a real, identifiable person without their genuine consent. Treat this as a higher-risk category and get a clear answer on consent and image rights before using it.

Is the ASA actually monitoring AI-generated ads specifically?

Yes. The ASA is scaling up its AI-based Active Ad Monitoring system in 2026 to proactively identify non-compliant advertising, including content generated or assisted by AI, rather than relying only on consumer complaints.

What happens if an AI-generated ad is found to be misleading?

The same enforcement process applies as for any other ad: the ASA can require the ad to be withdrawn or amended, and repeated or serious breaches can lead to further sanctions. The fact that AI produced the content is not a defence.