Why AI UGC doesn't convert — and the shot that fixes it
Every AI UGC ad looks the same: a girl holding a product, talking to the camera, smiling on cue. The format works when a real person does it, and collapses when a model generates it. That gap is not about model quality. It’s about direction.
Why it fails
The skin is too perfect. UGC sells because it looks unguarded. Porcelain skin with no pores, no stray hairs and no shine reads as an ad in half a second, and once it reads as an ad, the format has lost the only advantage it had. Ask for visible pores and no retouching, every single time.
There’s music. A real person filming in their bathroom has no soundtrack. The moment a track comes in, you’ve told the viewer this was produced. Room tone — water, fabric, footsteps, breath — is what makes it feel real, and it costs nothing.
The framing is too good. Centred subject, clean headroom, tidy background. Nobody films themselves like that. Slightly off-centre, a bit of handheld drift and an untidy shelf do more for credibility than another rendering pass.
The voice is performed. If it sounds like a voiceover artist, the spell breaks. UGC voice is close to the mic, a little quiet, slightly uneven.
The format nobody is using
There’s a structural problem underneath all of this: an AI person performing sincerity is uncanny, because sincerity is exactly what a model is worst at faking.
So don’t use one. Give the camera to the product.
Shoot the whole thing from the point of view of the object — a jar of cream on a bathroom shelf that watches someone walk in, get picked up, opened, used, and put back down. The person is in frame, but they’re not performing to camera. They’re doing something ordinary while being held.
It solves the credibility problem at the root: you can’t perform for a camera you’re holding without knowing it.
What it needs to work
Morning window light, no fill. Real bathroom, real shelf, a plant, a towel that isn’t folded. Handheld drift as the jar is lifted, so the frame tilts with the hand. And a voice — if you use one — written from the product’s side and whispered, never sold.
Then the last two seconds: the jar is placed back, the person walks out, and the room goes quiet. That silence is the whole ad.
The point
UGC is not a lower tier of production. It’s a different one, with its own rules, and most of those rules are about restraint: less polish, less music, less symmetry, less performance. The tools will happily give you more of all four. Knowing when to refuse them is the job.