When Everyone Has 2,500 Creatives, What Actually Becomes the Competitive Advantage?
AI has made one part of user acquisition significantly easier: creating ads.
And I think it has quietly made another part much harder.
Figuring out which ads actually matter.
AppsFlyer’s 2026 State of Gaming report recently put some numbers behind what many of us have already started to feel in the industry. Top gaming advertisers are now producing roughly 2,400–2,600 creative variations per quarter, up around 25–30% year over year. At the same time, ad impressions increased by 20%, while paid install share grew by 10%.
So we are producing more ads, putting more ads into the market, and competing for roughly the same human attention.
That made me think about something I have experienced from a slightly different angle.
I started my UA career working with playable ads, and today I work much more closely with rewarded traffic and multiple traffic sources. After working with 50+ sources, one thing has become increasingly clear to me:
A creative does not really have a performance on its own. It has a performance within an audience.
And I think this distinction is going to become much more important as AI makes creative production almost unlimited.
For a long time, the problem was production.
We needed more creatives.
More hooks. More videos. More gameplay variations. More UGC. More iterations.
Now AI can give us all of that at a speed that would have been difficult to imagine a few years ago.
AppsFlyer analyzed 1.1 million video creative variations across 1,300+ apps and found exactly this shift: creative production is scaling, but the bottleneck is moving from production toward measurement and optimization. In other words, making more creatives is becoming easier; understanding what to do with all of them is becoming harder.
And I think this changes the question we should be asking.
Instead of:
“Which creative is the winner?”
I would ask:
“Which creative is the winner, for which audience, on which source, and optimized toward what?”
That sounds like a small difference, but in rewarded traffic it can completely change how you look at performance.
I have seen the same game behave very differently across sources.
Sometimes a source has users who respond immediately to a strong reward proposition. The incentive is clear, the action is simple, and the conversion happens quickly.
Another source can have a very different audience. Users might respond much better to seeing the actual gameplay, understanding the progression, or knowing what they are going to achieve inside the game.
The game hasn’t changed.
The creative doesn’t necessarily have to change dramatically either.
But the user has.
And that changes everything after the install as well.
This is one of the reasons I have become a little skeptical of the idea that there is one universally “best” creative.
AppsFlyer’s creative research actually shows some very interesting examples of this. In gaming, the same type of creative outcome can perform very differently depending on the media environment. For example, “Pure Success” gameplay endings generated 33% higher IPM on ad networks, while “Pure Failure” performed 65% better on Social & Search. Even within UGC, Gameplay Reviews generated much higher IPM than Testimonials on ad networks, while Testimonials delivered stronger D7 retention.
To me, this is much more interesting than simply saying “Creative X performs better than Creative Y.”
It tells us that context matters.
And that is where my experience with rewarded traffic has changed the way I think about creative.
Rewarded is particularly interesting because the user already has a motivation beyond simply discovering a game.
There is an incentive involved.
So the creative is not just selling the game.
It is setting an expectation for the entire experience:
Why should I install this?
What do I get?
What do I need to do?
How difficult is it?
And is the reward worth it?
That means the creative, the incentive, the event and the traffic source are all connected.
If you change one, you can change the outcome of the others.
This is also why I don’t think we can talk about creative performance separately from traffic quality anymore.
Imagine two sources running the exact same creative.
Source A generates a very strong install rate but users disappear quickly.
Source B has a lower initial conversion rate, but the users who do install progress further into the game and retain better.
Which source has the better creative?
I don’t think we can answer that question without looking at what happens after the install.
Get Ecem Kır’s stories in your inbox
Join Medium for free to get updates from this writer.
This is something I also noticed much earlier while working with playable ads.
Playable ads taught me that the ad itself can become a first product experience. The user is not just watching a message; they are interacting with something that gives them an idea of what the actual product feels like.
Rewarded traffic adds another layer to that.
Now we are not only thinking about whether the ad communicates the game correctly. We are also thinking about whether the promise made by the ad, the reward and the actual user journey are aligned.
That alignment matters.
If the creative creates one expectation, the reward creates another, and the game delivers something completely different, you may still get the install.
But you may not get the user.
And I think this is where the industry is gradually moving from optimizing for acquisition metrics to optimizing for the entire acquisition experience.
Adjust’s 2026 gaming research points in a similar direction, with more emphasis on retention, high-LTV players, stronger engagement loops and longer player lifecycles rather than simply maximizing scale.
Rewarded UA is moving in the same direction.
Singular’s 2026 analysis argues that rewarded acquisition is increasingly becoming an LTV-first channel, with longer measurement windows and more focus on retention-adjusted value rather than only CPI or early ROAS. It also highlights how AI could eventually personalize not just creative variations, but the reward path itself.
That is where I think things get really interesting.
Because if AI can eventually generate hundreds of creative variations, and potentially adapt messaging based on user behavior, the competitive advantage won’t be having access to AI.
Everyone will have it.
The advantage will be knowing what to ask AI to create in the first place.
And that knowledge comes from understanding users.
This is probably the biggest lesson I have taken from managing so many different sources.
Every source has its own behavior.
Some sources are extremely strong at driving initial engagement.
Some are better at deeper progression.
Some audiences are much more reward-sensitive.
Some respond better to gameplay.
Some can scale very quickly but become less efficient as volume increases.
Some look average at D1 and become surprisingly strong later in the cohort.
Once you work across enough sources, you stop seeing them as just lines in a dashboard.
You start seeing patterns.
And those patterns should influence creative strategy.
Maybe the question isn’t:
“What creative should we run?”
Maybe it’s:
“What does this audience need to see before they are willing to take the next step?”
That is a very different way of thinking about creative. It also changes how we should measure it. I don’t think CTR or IPM suddenly became irrelevant. They are still incredibly useful. But a high IPM creative that brings low-quality users isn’t necessarily a winner. Likewise, a creative with lower initial engagement can sometimes bring a much more valuable cohort.
AppsFlyer’s own creative research makes this point by looking at share of cost, IPM and D7 retention together. A creative with relatively low spend can still have strong IPM and retention because it resonates with a smaller, higher-value audience.
That is particularly relevant for rewarded traffic. Because when you are looking at the performance of a source, you eventually need to connect:
creative → install → event → progression → retention → LTV
And I would add one more layer:
source.
Because the same creative can produce a completely different journey depending on where the user came from.
This is also why I think the future of creative optimization is going to look less like a creative team’s problem and more like a cross-functional UA problem.
Creative teams can tell us what we can make.
UA teams can tell us what converts.
Traffic managers can tell us how different sources behave.
Product teams can tell us which user actions actually matter.
And the interesting part happens when those pieces come together.
When I look at the current direction of the industry, I don’t think AI is going to make creative less important.
Quite the opposite.
It is going to make creative intelligence more important. When everyone can produce 2,500 variations, producing the 2,501st variation isn’t necessarily where the advantage is. The advantage is knowing why one variation works on one source and fails on another. Knowing when a reward should be the hero of the message and when it should be secondary. Knowing whether you should optimize for an install, a progression event, or a deeper user action. Knowing when a source needs more volume and when scaling it further will simply dilute quality. And, perhaps most importantly, knowing when a creative is actually bringing the right user rather than just bringing more users.
So I don’t think AI has solved the creative problem.
I think it has changed the problem.
We used to struggle with creative scarcity.
Now we are moving into an era of creative abundance. And abundance creates a different challenge: deciding what deserves attention. For me, that is where traffic quality becomes much more important. Because when the number of creatives becomes almost infinite, the differentiator won’t be who can make the most ads.
It will be who understands the relationship between the creative, the audience, the incentive, the source and the outcome better than everyone else.
And after spending years on the creative side with playable ads and now managing rewarded traffic across 50+ sources, I think that is where the next phase of UA is going.
AI can create more ads.
But it still takes people, data and experience to understand which users those ads should actually reach.
