Can you use AI models in beauty advertising?
Yes. There is no UK rule against AI generated models in beauty ads, but the ad must not mislead about what the product does, and you must never use a real person's likeness without their written permission.
Is it legal to use AI models in UK beauty ads?
Yes. At the time of writing there is no UK law or CAP Code rule that bans AI generated people in advertising. The Advertising Standards Authority (ASA) has said its existing rules apply to ads however they were made. So an AI model sits under the same rules as a photographed one.
That cuts both ways. You don't need special permission to use a generated model. But you also can't use AI as an excuse. If an ad misleads, the advertiser is responsible, even if a tool produced the image.
What does the ASA actually care about?
The core question is simple: does the ad give a misleading impression of the product? The ASA has pointed out that AI generated images could mislead if they don't reflect what the product can really do, in the same way that heavy retouching or social media filters can.
Beauty has a clear precedent here. In 2021 the ASA ruled that filters should not be applied to images promoting beauty products if they are likely to exaggerate the effect the product can achieve. That thinking carries straight over to AI. A generated model with flawless, poreless skin in a foundation ad raises the same question as a filtered influencer.
- Efficacy: does the image show a result the product can't deliver, such as lash length, skin smoothness, hair volume or wrinkle reduction?
- Substantiation: if the ad makes a claim, can you back it up with evidence, regardless of what the picture shows?
- Social responsibility: does the image promote an unhealthy or unrealistic body image, or target under 18s inappropriately?
- Testimonials: does a generated face imply a real customer who used the product?
Where do AI models go wrong in beauty ads?
Most problems come from using a generated face to demonstrate a result. If your mascara ad shows an AI model with dramatic lashes, the viewer reads that as what the mascara does. If the lashes were generated rather than created by the product, that is a misleading efficacy claim.
The safer approach is to separate mood from proof. Use AI models for atmosphere, styling, campaign worlds and showing who the product is for. Use real, honest evidence for anything that shows a result. Before and after imagery in particular should only ever come from real, unaltered results. Generating a treatment result is a fast route to an upheld complaint.
| How the AI model is used | Risk level | Why |
|---|---|---|
| Campaign mood: model holding the product in a styled scene | Lower | Shows context, not a result |
| Model wearing a lipstick shade | Medium | Shade must match the real product exactly |
| Model showing smooth skin in a foundation or serum ad | Higher | Viewers read the skin as the product's effect |
| Generated before and after results | Very high | Fabricated efficacy evidence |
| Face resembling a real person or celebrity without consent | Very high | Likeness, passing off and data protection issues |
Do you need consent to use an AI model?
If the face is original and doesn't resemble anyone real, there is no individual whose consent you need. The problem starts when a generated face is based on, or looks like, a real person. That includes training or prompting a model on a specific creator's photos, or making a lookalike of a celebrity.
In the UK, using someone's recognisable likeness to suggest they endorse your brand can lead to a passing off claim. A person's image is also personal data under UK GDPR. On top of that, the CAP Code says you shouldn't feature or refer to people in a way that implies endorsement without their permission. The simple rule: if a real person is recognisable, get their written permission, with clear terms on usage, channels and duration, just as you would for a model booking.
A good studio will hold to this as standard. At Niko x Studios, real people are only featured when the brand holds their written permission. Otherwise the studio directs original talent that isn't based on anyone real.
Do you have to label an AI model as AI?
Not always under the CAP Code. There is no blanket rule that every AI image must carry a label. But if leaving it out would mislead people, for example if viewers would assume a generated person is a real customer sharing a real result, then you need to make the truth clear.
Platforms set their own rules, and they change often. Meta and TikTok both have policies on AI generated or digitally altered content, and some ad types need a disclosure. Check each platform's current advertising policy before you launch. For the full picture across the UK, Meta, TikTok and Google, see our guide on whether you have to label AI ads.
How do you use AI models honestly?
Treat a generated model like a cast member with a brief. Decide what job the image is doing before you make it. If the job is to show a result, the result has to be real.
Representation matters too. AI makes it easy to show a wide range of skin tones, ages and hair types. Use that, but make sure the product shown on each person matches what it would genuinely look like on them. A foundation shade that looks perfect on every generated skin tone, when the real range doesn't cover them, is its own kind of misleading.
- Keep the product itself accurate: shade, packaging, label and texture checked against the real thing.
- Don't use AI to show a result. Use real, unretouched evidence for efficacy.
- Avoid flawless, poreless skin in complexion ads. Realistic skin texture is safer and often more believable.
- Don't present a generated person as a real customer or reviewer.
- Get written permission for any recognisable real person.
- Keep notes on how each asset was made, so you can answer a complaint quickly.
What should you ask a studio before using AI models?
Ask how they create faces and whether any are based on real people. Ask how they check the product against the real thing, and what they won't make. A studio that will happily generate before and after results is a risk to your brand, not a shortcut.
You should also be clear on who holds the rights to the final images and the generated talent. Our guide to who owns AI generated images covers this, and you can see how campaign visuals with directed talent come together on our AI campaign visuals page.
This guide is general information, not legal advice.
Questions people also ask
Can I use an AI model in a skincare ad?
Yes, as long as the image doesn't suggest the product achieves a result it can't. Use the model for mood and context, and keep any efficacy claims backed by real evidence.
Can I make an AI version of an influencer I work with?
Only with their written permission, covering how, where and for how long the likeness is used. Without it you risk passing off and data protection issues, and the ad may imply an endorsement they never gave.
Has the ASA banned AI in beauty ads?
No. At the time of writing the ASA applies its existing rules to AI generated ads. The main concern is misleading efficacy, the same issue it raised with beauty filters.
Are AI generated before and after photos allowed?
They would almost certainly mislead, because they show a result that never happened. Before and after images should only use real, unaltered results.
Do AI models have to be diverse?
There is no rule requiring it, but showing a range of people is common good practice. Just make sure the product looks accurate on each person shown.
Tell us the product and we'll send a free campaign image of it within 5 working days.