Decide what survives the next version
A successful ad gives you several things you could change: the hook, product reveal, demonstration, proof, pacing or offer. Changing them all makes the next result difficult to interpret and can remove the reason the original was useful.
A good iteration recommendation names what to preserve before explaining what to change. For example, keep the demonstration that makes the product understandable, then test a more specific opening that gets the right buyer into that demonstration. This is an illustrative design principle, not a claim about a measured campaign uplift.
Score the recommendation separately
Benmark asks for the analysis and the creative action as separate outputs. The model that writes the strongest explanation is therefore not automatically treated as the best model for the next idea.
Claude Opus 5 leads the reviewed iteration score, while Google Gemini 4 Argon shares the leading creative-analysis score. The separation is useful when designing a workflow: diagnosis and development can be assigned according to the task results rather than a single overall model ranking.
Make the next test interpretable
Ask for one recommended route, the element it preserves, the change it makes and the response you expect that change to affect. A production team should be able to act on the recommendation without having to invent its commercial logic.
The test supplies the ad itself without account performance. When you use this in practice, add the relevant performance breakdowns so the model can distinguish a promising creative mechanism from an attractive explanation of an ad whose result came from something else.
How Benmark tests this
Using five image ads and five video ads, models recommend how to develop a stronger next version. Ben scores the recommendation separately from the explanation of the original ad.
Test inputs, instructions and scoring
What the model receives
The fixed pack contains five static images and five videos. Models see the media, with original filenames and account performance withheld. They receive no spend, conversion, reach or brand-context evidence. Analysis and creative recommendations are generated as separate outputs.
What changes between prompts
Each asset is tested with a simple and a scaffolded prompt. The analysis scaffold covers attention, product relevance, proof, platform fit and other creative mechanisms. The action scaffold asks what to preserve, what to change and why. The media remains identical between conditions.
How the result is scored
Action outputs are judged on the recommendation's creative and commercial quality, including coherence/commercial judgment as a critical dimension. These are ratings of the proposed next version. Neither the generated ad nor its subsequent campaign performance is tested in this task.