lizyarova
9月 10, 2026
Generative AI has changed how quickly growth teams can produce mobile ad creatives. New hooks, voiceovers, visuals, and full ad variations can now be created in minutes rather than days or weeks.
Still, producing more creatives does not automatically lead to stronger performance. Despite the advances in AI, every variation requires testing, time, budget, and reliable data. And, teams need enough evidence to decide whether to scale, revise, or stop a creative.
In a recent episode of Tenjin’s ROI 101 podcast, Roman from Tenjin speaks with Mark, VP of Growth at Bidmatrix, about how this shift is changing creative testing and decision-making. His central point:
“We are no longer short of ideas. We are short of reliable decisions.”
AI Shifted the Production Bottleneck
Before AI became widely available, a new creative round often meant collaborating with designers, editors, or video producers and waiting for space in the production calendar.
Every creative concept competed for limited time and resources. That naturally limited how many ideas a team could produce and test. But, with AI workflows, many of those constraints have been lifted, allowing creative teams to generate multiple hooks, visual treatments, voiceovers, and localized versions in a single day.
While this has driven an increase in creative content production, Mark argues that it has also shifted the bottleneck elsewhere:
“The new bottleneck is judging ads, not making them.”
He argues that creative production has accelerated, but testing capacity has not grown at the same rate. Budgets remain limited, meaningful results take time, and not every variation receives enough data to support a confident decision.
More Creatives Create More Testing Debt
Cheaper, faster production gives teams more options. The challenge is deciding which variations deserve testing without spreading budget and attention too thinly.
Mark describes this problem as testing debt:
“Teams can create tens or hundreds of options in a day. That sounds positive, but it can create testing debt, with more ads than the team can properly test. When the budget is spread across too many small variations, the results become noisy and false winners can appear.”
Testing debt is the growing backlog of creatives that a team can produce but cannot evaluate reliably.
Even with a larger budget, teams cannot test thousands of variations effectively. Every ad still needs to demonstrate real economic value before it earns more investment.
This is why more creative volume does not always produce better learning or stronger performance.
Ten genuinely different ideas can quickly become 100 versions with only small changes to a background color, a line of copy, or a visual detail. AI makes those versions easier to produce, but having more variations does not automatically make them all worth testing.
Three Questions for Creative Accountability
If creative output is no longer the best measure of progress, growth teams need a better way to judge performance.
Mark frames creative accountability around three questions:
- What did the ad promise?
- Who did it bring in?
- Did those users stay and create value?
Together, these questions connect each ad to the quality of the users it attracts.
Mark explains that this downstream focus is central to how Bidmatrix approaches optimization:
“We optimize toward downstream events because we are interested in what happens after an ad gets attention.”
Success is not only measured by a click or install. A successful creative attracts users whose experience matches the promise in the ad and who go on to engage, retain, or create economic value.
AI can make it easier to produce dramatic hooks and attention-grabbing concepts. However, a creative may perform well at the top of the funnel while creating a poor match between the ad promise and the actual product experience.
The Creative Darwinism Loop
Mark turns creative accountability into a repeatable framework called the Creative Darwinism Loop.
The process is:
Research → Generate hypotheses → Filter → Test → Validate → Scale carefully
“Ads have to earn the right to survive. Many ideas enter, but only ads that prove value should receive more budget.”
Teams begin by researching what users care about and turning those insights into a small number of clear hypotheses.
They then remove ideas that are confusing, off-brand, irrelevant, or likely to create a mismatch with the product.
The remaining creatives receive controlled test budgets. Only after the team validates both initial appeal and downstream value should an ad receive more spend.
AI can support every stage of this process. It can summarize user language, identify competitor angles, suggest hypotheses, generate variations, and help analyze results. However, it should support the process rather than replace it. The goal is not to automate every judgment or create the largest possible folder of creative assets.
“AI should strengthen the loop…with the goal to learn faster from the market.”
Why Testing Still Requires Human Ownership
AI is useful for research, ideation, production, and creative analysis. But, testing and validation still require human eyes and ownership because those stages involve real budgets and limited data. Inferred decisions should have someone supporting them. As Mark explains:
“Human input matters most during testing and validation, because that is where real budgets and real decisions are involved.”
An AI model can summarize early performance metrics, identify patterns, or propose a creative direction. People still should review the setup and decide whether a campaign has earned more resources. The closer a decision gets to real spending, the more important it is that someone steers the judgement.
Keep the Hook Aligned with the Product
AI is particularly useful for generating opening hooks designed to stop the scroll. However, a person must still check whether the hook’s promise matches the actual product.
The core ad should demonstrate the real product experience or authentic gameplay. Without this, teams risk scaling misleading winners. This includes ads that generate strong click-through rates but attract users who churn quickly and weaken overall unit economics.
Human judgment also matters when deciding which emotional promise fits the product. Sure, an AI can suggest angles based on relief, confidence, urgency, or control, but the sentience needs to match what the app actually provides.
This does not mean every decision must remain manual forever. Automation should be always be introduced after the guidelines have been established: the team has defined the right checks, data, and decision rules.
Why Measurement Matters More in AI Creative Testing
As AI increases creative production capacity, it becomes easier for teams to confuse top-of-funnel attention with long-term business value. This is why measurement becomes more important, not less. In order to valuate ads accurately, growth teams need to distinguish between two layers of performance data.
Creative Appeal
CTR and IPM indicate whether an ad attracted attention and generated installs. These metrics help answer questions such as:
- Did the hook stop the scroll?
- Did users click?
- Did the ad generate installs efficiently?
User Quality
Retention, key in-app events, ROAS, and LTV indicate whether those users stayed and created economic value. These metrics help answer:
- Did users engage with the product?
- Did they return?
- Did they complete meaningful actions?
- Did the campaign make business sense?
Mark separates these metrics into three groups:
“CTR and IPM tell us whether the ad received attention and installs. Retention tells us whether people were interested in the app. ROAS and LTV tell us whether the campaign made business sense.”
Both layers matter. The challenge is that appeal metrics arrive first, while downstream user quality can take days or weeks to observe. That delay can make an early winner appear more conclusive than it really is, leading to a costly scaling mistake.
“The most dangerous ad is one that looks like a winner too early.”
AI can quickly generate dramatic, attention-grabbing hooks. This also makes it easier to create ads that drive inexpensive clicks but attract users who churn soon after installing. The faster a team can produce and test variations, the more disciplined it needs to be about evaluating both appeal and quality before scaling budget. As Mark puts it:
“An ad should pass both gates before receiving any major budget.”
Decisions Drive AI Creative Testing
Generative AI has expanded what growth teams can produce. However, it has not removed the need for clear hypotheses, controlled tests, reliable measurement, or human accountability.
When every team can generate more creatives, the advantage will come from knowing what deserves to be tested, what deserves more budget, and what should inform the next creative decision.
You can always watch the full ROI 101 conversation with Mark to learn more about testing debt, creative accountability, and the role of AI across the creative process.
