Adding a multi-language field to the backend takes an afternoon. The part that takes months sits in …
100 Creatives in Minutes. The Landing Page Takes Three Weeks.
100 Creatives in Minutes. The Landing Page Takes Three Weeks.
System1's new "Test Your Ad Screen" tool checks up to 100 finished creatives at once and delivers a result in minutes, work that used to take weeks. The more interesting part of the announcement isn't the speed, though. The model refuses to give an answer when its prediction is uncertain instead of forcing out a number anyway. And that new speed exposes exactly where the next bottleneck sits: no longer with the creative, but with the page it points to.
What System1's "Test Your Ad Screen" actually delivers
According to a report by Meedia, System1's new tool screens up to 100 finished long-form creatives simultaneously. The underlying model is trained on 18 million real emotional reactions to advertising. Instead of a single score, the tool outputs low, medium, or high confidence bands, and when prediction certainty is too low, it returns no result at all. System1 itself frames the tool as complementing human research earlier in the process, not replacing it. It is currently available only in the US and the UK.
The methodologically interesting part: when the tool stays silent
The sentence that's easy to miss in the announcement is the one about confidence bands. A model that refuses to answer instead of producing a number regardless is methodologically cleaner than most AI tools, which tend to output some value rather than none. Anyone who has seen a dashboard project false certainty knows how rare that restraint is.
But that clean handling of uncertainty only shifts the problem, it doesn't solve it. When creative screening drops from weeks to minutes, testing stops being the bottleneck in the campaign process. What comes next becomes the bottleneck instead.
The bottleneck moves: from the creative to the landing page
100 pre-screened creatives don't get you far if every matching landing page is a dev-team ticket with a three-week lead time. That's the daily reality in many marketing teams today: the creative is validated in minutes, and the landing page it needs still requires an engineering ticket, a sprint slot, and an approval loop. The velocity gap you used to look for in the creative now sits with the page.
That's the same pattern showing up as Google merges Tag and Tag Manager: a tool gets faster, but the frontend it's aimed at stays the actual bottleneck. Testing without the ability to build means testing against a wall.
The direct fix is a frontend where marketing teams can assemble the matching landing page themselves, without waiting on a dev ticket. A composable visual page builder turns a validated creative into a live landing page in hours instead of weeks, and an A/B test runs from the same editor without engineering having to release the variant first. It's the same idea behind the velocity gap between marketing and frontend autonomy: autonomy only helps if it sits in the right place.
Two caveats that belong in this story
Editorial honesty means not skipping two points here.
First, the widely quoted "nine out of ten cases" match with human testing is a claim from System1 itself. It is not an independent study and not an externally validated benchmark. That doesn't make the number worthless, but it needs to be cited with that context, not treated as an objective metric.
Second, a model trained on past data inevitably learns what worked in the past. That means a structural bias against genuinely new work, against formats and visual language that weren't part of the 18 million training reactions. If you're testing a creative that breaks with convention, keep that in mind before you trust a low confidence band.
What this means for product/marketing owners and creative leads
| Dimension | Before | With a composable frontend | |---|---|---| | Creative screening | Weeks, often outsourced | Minutes, AI-assisted pre-sort | | Landing page for the creative | Agency or dev ticket, roughly three weeks | Hours in the editor, no ticket | | A/B testing the variant | Waits on deployment | Runs from the same editor | | Who controls the pace | Engineering capacity | The marketing team itself |
If your creative screening gets faster but your landing pages don't keep up, the speed gain evaporates exactly where it matters: at the click.
The same handoff gap shows up when 70 percent of DACH online shoppers order straight from a newsletter but the landing page behind it stays generic: the purchase impulse is there, the page just isn't built fast enough to capture it.
FAQ
Does AI creative testing replace human user research? No. System1 itself positions the tool as an early-stage complement in the process, not a replacement for human studies.
Is the "nine out of ten cases" figure a solid benchmark? It's a vendor claim from System1, not an independently validated study. A reliable assessment would need external verification that doesn't currently exist.
Why does the tool sometimes return no result at all? When prediction certainty is too low, the model withholds a confidence band instead of presenting an uncertain number. That's the methodologically correct call, even when it's inconvenient in practice.
What does the bias against new work mean in practice? The model scores creatives based on past emotional reactions. Genuinely new formats or visual language aren't represented in that training data and tend to score worse, regardless of whether they'd actually work with a real audience.
How do you close the velocity gap between creative and landing page? By letting marketing teams build and test the landing page themselves in an editor, instead of opening a developer ticket for every variant.
The automation keeps moving, the question stays the same
Google rewrites product titles, builds events by click, and now an AI pre-sorts in minutes what used to take teams weeks. The common thread is always the same: the faster the layer above the frontend gets, the more visible it becomes whether the frontend itself can keep up. For System1's creative screening, the answer to that question is the landing page, and it decides whether those three minutes of testing were worth anything at all.
If you want to see how landing pages get built at campaign speed instead of sprint speed, start with A/B testing without a dev ticket.
About the author: The Laioutr Team writes about frontend management, composable commerce, and how marketing teams take back control of their frontend.