We Tested AI-Generated Ad Creative Against Human-Made, Results Were Mixed

We ran a controlled test: the same campaign brief given to a human creative team and to an AI image and copy generation pipeline, both sets of creative run head-to-head in the same ad sets. Neither side won cleanly, and the pattern of where each won told us more than an overall win rate would have.

Where AI creative won

For static product-focused ads with clear, literal briefs — show the product, state the offer, use the brand colors — the AI-generated creative performed statistically on par with the human-made set, at a fraction of the production time and cost. For rapid iteration across many small audience segments, where volume mattered more than any single ad's polish, AI creative let us test variations the human team simply didn't have the hours to produce.

AI-generated creative isn't competing with your best human creative. It's competing with the creative you'd never have had the budget to make at all.

Where it lost, clearly

• Any creative requiring genuine brand voice nuance or humor landed noticeably flatter and performed worse — AI-generated copy tended toward generically upbeat phrasing that didn't match the brand's actual tone.
• Video and motion creative was not close — the human team's video ads substantially outperformed AI-generated video across every metric we tracked.
• Ads targeting a specific cultural or regional nuance (a local reference, a specific community's shorthand) consistently underperformed when AI-generated, since the generation models had no access to context that specific.
• Audience fatigue set in faster on AI-generated creative sets, possibly because the underlying generation patterns produced visually similar outputs across variations even when we asked for different concepts.

The useful takeaway wasn't 'AI creative works' or 'AI creative doesn't work' — it was a map of where each approach has the edge. We now default to AI for high-volume static iteration and keep the human team on video, brand-voice copy, and anything culturally specific, which is a more precise allocation of budget than either extreme position the industry debate usually argues for.