AI UGC ads, and where they beat a real creator
AI UGC ads work when the job is volume, consistency and speed, and they fall apart when the job is trust in a specific person. The useful question is not whether they look real, but whether the claim being made depends on who is making it.
AI UGC has gone from novelty to default fast, and the debate has mostly been about whether viewers can tell. That is the wrong question. Plenty of ads that are obviously not real people work extremely well.
The better question is what the ad is asking the viewer to believe, and whether that belief depends on a human being behind the claim.
What we mean by AI UGC
Two different things get called the same name. One is a generated person delivering a script in a way that imitates a filmed testimonial. The other is a built character with a consistent face, voice and world, used the way a brand mascot is used.
We build characters rather than cast people. A claymation figure, a talking object, a crochet build: designed and produced in house, so the same character can carry a campaign across a market without scheduling anyone.
That distinction matters for what you can honestly claim. A built character can hold attention and explain a product. It cannot vouch for one.
Where AI UGC ads win
- Volume. A character can carry twenty variations without twenty shoot days, which is what makes weekly testing possible.
- Consistency across markets. The same character delivers the same idea in four languages without four castings.
- Novelty against ad blindness. A format the feed has no reference point for buys back the first three seconds.
- Anything awkward to film. Cutaways, impossible objects, scale changes, the inside of a product.
- Speed. When the replacement has to exist before the current winner fades, a built character has no calendar.
The novelty case is the strongest and the most often missed. Lead Printer had sat between 40 and 50 euros per qualified meeting booked for as long as their CEO could remember. We sent a claymation concept, which nobody expects from a print business.
I did not think it would work for a business like ours. It came in at 9 euros.
Where it falls apart
The failure is rarely visual. It is a mismatch between the format and the claim.
- Personal testimony. If the line is I used this for three months and here is what happened, the ad needs a person who actually did.
- Categories where trust is the purchase. Health, finance, anything involving a body or a bank account.
- Before and after results. A generated result is a claim about a thing that did not happen.
- Expertise claims. A built character asserting professional authority reads as costume, because it is.
There is also a regulatory line worth respecting rather than testing. An endorsement implies a real person with a real experience, and a character cannot supply one.
If the problem is a trust deficit, change the structure
When people watch but do not click, or click but do not buy, the account usually has a belief problem rather than an attention problem. AI UGC is a poor fix for that, and a different structure is a good one.
A podcast-style slice works because the proof arrives as a conversation between two other people rather than a message pointed at the viewer. One of ours ran at $0.33 CPC.
Talking head with B-roll works for the same reason. The format signals that someone is accounting for what they say.
How to use built characters without losing credibility
- 1Decide what the ad asks the viewer to believe before choosing the format.
- 2Give claims that need a person to a person, and give demonstration, novelty and explanation to a character.
- 3Keep the character consistent, so repeated exposure builds recognition rather than confusion.
- 4Do not dress a character as a customer. Ambiguity about who is speaking is what damages trust, not the fact that it is animated.
- 5Test it as its own concept with its own hypothesis, rather than as a cheaper version of a creator ad.
The real advantage is supply, not cost
AI UGC gets sold on price, which undersells it. The advantage is that a built character removes the scheduling constraint from creative production.
That is what makes a fixed cycle possible: the next test can exist before the current one fades, because nothing waits on a calendar. Cheap ads that arrive too late still cost you the account's momentum.
Questions we get asked
- What are AI UGC ads?
- Ads that use a generated or built character in place of a filmed creator. That ranges from a synthetic person delivering a script to a designed character such as a claymation figure or a talking object used consistently across a campaign.
- Do AI UGC ads perform as well as real creator ads?
- For attention, explanation and novelty, often yes. For personal testimony they do not, because the claim depends on a real person having had the experience. Match the format to what the ad asks the viewer to believe.
- Are AI UGC ads allowed on Meta?
- Generated creative is permitted, but the usual advertising rules still apply: no misleading claims, and endorsements must reflect genuine experience. Presenting a character as a real customer is the line worth staying well clear of.
- Do you need to disclose AI in ads?
- Requirements differ by market and are tightening. The practical rule is to avoid ambiguity about who is speaking, since a character that is clearly a character raises no question in the first place.
- Do you source creators?
- No. We design and build creator characters in house, which is how one character can carry a campaign across markets without casting or scheduling anyone.