Two numbers worth knowing
Nearly two in three video ad buyers in the US now use generative AI in their creative, up from half a year earlier, according to IAB. A third of their ad assets will use it this year, rising to 43% by 2027.
When Kantar studied hundreds of ads made with generative AI, the ones where the AI was obvious scored lower on branding. The ones where it was seamless did well: more than 40% of them landed in the top tier for branded cut-through.
What the film shows
We made this 15-second film to explain that finding. Halfway through, two frames sit side by side. One has a tell, marked in red: a hand that doesn't look quite right. The other has none. That gap is what audiences react to, and it's the difference Kantar measured.
What gives AI away
- A face that shifts slightly from one shot to the next.
- A product that changes shape, colour or detail as the camera moves.
- Light and reflections that don't match the room.
What we hold fixed
- The product's details, checked frame by frame against the original photos.
- One face, designed once and held consistent through a whole film or a whole catalogue.
- Light that belongs to the scene, so nothing looks pasted in.
Why it matters for brands
Kantar's finding is easy to read as a reason to hide AI. We read it as a standard. Audiences don't mark a film down for how it was made. They mark it down when something looks wrong: a face that drifts between shots, a product that changes shape, light that doesn't belong to the room.
Generating a beautiful scene is the easy part now. Keeping a real product identical while everything around it moves is what decides whether anyone notices. Most ad creative now involves AI, so using it is no longer an advantage. Making work nobody can spot as AI, at volume and every single time, is.
Sources: Kantar, Rethinking AI-generated advertising, November 2025. IAB, 2026 Digital Video Ad Spend & Strategy Report, July 2026.