UGC ads changed paid social because they did not look like ads. A person holds a phone, talks about a product in their kitchen, and the viewer stays for the first three seconds because it feels like something a friend posted. The problem has always been supply. Finding creators, shipping products, waiting for footage and asking for reshoots takes weeks.
AI UGC removes the waiting. The person, the room, the voice and the phone-camera look are generated. That makes some things much easier, and it creates a few problems that are worth understanding before you spend money.
What AI UGC ads actually are
An AI UGC ad is a vertical video in the familiar creator style, where the presenter is generated. There are two main ways to make one. The first uses an avatar system: you pick or design a presenter, type a script, and the system produces a talking video. The second is built shot by shot, the way a studio works: a designed character, generated scenes and a recorded or synthetic voice, all cut together like a normal ad.
The first method is fast and cheap, and everyone’s ads look the same. The second takes longer and looks like your brand.
Where they work
Testing hooks. The first three seconds decide whether an ad gets watched. With filmed creators, testing twelve different openings means twelve sets of footage. With AI it means an afternoon. This is the strongest argument for the format.
Simple, clear offers. Apps, subscriptions, e-commerce bundles and seasonal promotions. If the message is “here is the problem, here is the product, here is the discount”, the delivery matters less than the clarity.
Explainers and how-tos. A consistent presenter who walks through a feature works well, and it is easy to update when the product changes.
More than one language. The same presenter can deliver the same ad in English and Russian with matching lip movement, which is hard and expensive to do with real creators.
Where they backfire
When trust in a person is the product. Skincare results, fitness transformations, financial advice, medical services. Viewers are asking “did this really work for someone like me?” and a generated person cannot honestly answer that.
When the presenter pretends to be a customer. This is the line that matters. An AI presenter saying “here is how this app works” is a presenter. An AI presenter saying “I lost ten kilos with this” is a fabricated testimonial. Advertising rules in the UK require that ads do not mislead, and the US Federal Trade Commission has rules against fake reviews and testimonials, including ones made with AI. Platforms can reject the ad and regulators can fine the advertiser.
When the audience spots it and feels tricked. Viewers are getting better at noticing generated people. If the ad is honest about what it is, most people do not mind. If it pretends to be a real customer and gets caught, the comments will say so.
What the advertising rules say about AI testimonials
The short version: a generated presenter is allowed, a generated customer is not.
In the UK. The CAP Code, which the Advertising Standards Authority enforces, says marketing must not materially mislead (rule 3.1) and must not contain fake consumer reviews (rule 3.44). Marketers must hold documentary evidence that a testimonial or endorsement is genuine, plus contact details for the person who gave it, unless it is obviously fictitious (rule 3.47). The ASA’s advice on testimonials and endorsements explains what counts as evidence, and an invented person cannot supply any.
The law backs this up. The Digital Markets, Competition and Consumers Act 2024 made fake reviews a banned practice from 6 April 2025. It defines a fake consumer review as one that “purports to be, but is not, based on a person’s genuine experience”, and it bans both submitting one and commissioning someone else to write it. The Competition and Markets Authority has published guidance on fake reviews and can fine a company up to 10% of its global turnover.
In the US. The FTC’s Rule on the Use of Consumer Reviews and Testimonials took effect on 21 October 2024. It bans fake or false consumer reviews and testimonials, including ones attributed to someone who does not exist, and the FTC names AI-generated fake reviews as an example. Courts can impose civil penalties for knowing violations. The FTC’s questions and answers on the rule deal with AI avatars directly: there is no blanket ban on using them in marketing, and an avatar’s testimonial breaks the rule only if the underlying testimonial is fake or false, although it can still be deceptive under the FTC Act. The older Endorsement Guides apply as well: an endorsement must reflect the endorser’s honest opinion, nobody may describe an experience with a product they have not tried, and paid connections must be disclosed.
Here is how that sorts typical script lines:
| Line in the script | Verdict |
|---|---|
| “Here is how the app works”, followed by a demo | Fine. A presenter is explaining. |
| “Thousands of people already use this” | Only if you can prove the number. |
| “I used this for a month and my skin cleared up” | A fabricated testimonial. Do not run it. |
| A real customer’s review, quoted on screen | Only if the review is genuine, you hold the evidence, and it is clear who is speaking. |
| An AI lookalike of a celebrity praising the product | Not without the celebrity’s permission. |
This is general information, not legal advice.
How to use them well
- Write the presenter as a presenter. Let them explain, demonstrate and recommend. Keep customer claims for real customers.
- Follow the labelling rules. Meta and TikTok both have policies on labelling realistic AI-generated content. They change, so check the current version before a campaign.
- Design your own character. A presenter made for your brand is recognisable across ads. A stock avatar is recognisable too, as the same face selling a competitor’s product.
- Test one thing at a time. Same offer, same body, different hook. Then same hook, different offer.
- Mix formats. Use real customer clips where you need proof and AI variations where you need volume. The accounts that do best usually run both.
A worked example: one offer, six openings
This is an invented example, not a client result. Imagine a language-learning app selling an annual plan at a discount. The body of the ad stays the same in every version: fifteen seconds showing a lesson, the price and the button. Only the first three seconds change.
- The problem. “You have been ‘learning Spanish’ for three years and still cannot order dinner.”
- The question. “What can you really learn in ten minutes a day?”
- The contrarian line. “Stop memorising vocabulary lists.”
- The demo first. No speech at all: the screen, a finger, a lesson already in progress.
- The objection. “I thought I was too old to start a language.” This one belongs to a real user, so either film a real user or cut it.
- The offer. The discount in the first second, the explanation afterwards.
Five of the six can be made honestly with an AI presenter. Number five is a customer’s story, and it shows where the format stops. Run the versions against each other with the same audience and budget, then judge them on how many viewers stay past the first three seconds and on what each result costs, not on likes.
What to expect from a studio
A set of AI ads should arrive as a package: one agreed look, six to twenty versions with different openings, exports in 9:16, 1:1 and 16:9, captions and end cards, and file names your media buyer can upload without asking questions. If you also want a lasting face for your brand, that is a separate piece of work: see AI presenters.
At Zero Shot, an AI video studio in London, we write the presenter’s script as a presenter’s script and flag any line that reads like a customer’s claim before production starts.
If you are weighing this against filming real creators, our guide on choosing between AI video and a real shoot covers the trade-offs. If you are comparing suppliers, read how to choose an AI video production company first.
To get a fixed price for a set of ad variations, send us a short brief.
Questions people ask
What does AI UGC mean?
UGC stands for user-generated content: informal, phone-style video that looks like it came from a customer or creator. AI UGC recreates that style with a generated presenter, voice and setting instead of a filmed person.
Are AI UGC ads allowed on Meta and TikTok?
Yes. Both platforms accept AI-generated ads, and both have rules about labelling realistic AI content. The ad must still follow normal advertising rules, which means no invented customer reviews or misleading claims.
Do AI UGC ads perform as well as real creators?
It depends on the product. For simple offers, a strong hook matters more than who delivers it, so AI versions can match filmed ones. For products that depend on personal trust, real people usually win.
How many variations should I test?
Start with one offer and six to twelve different openings. Keep the rest of the ad identical so you learn which hook works, then make more versions of the winner.