The State of Creative Optimization – 2026 Edition

The State of Creative Optimization - 2026 Edition
01 KEY findings
3,727
Creative variation counts jumped 46% YoY at the $7M+ quarterly spend tier
As AI usage becomes mainstream, the surge in production concentrates almost entirely above $1 million per app in UA spend: every tier below it flattened or pulled back in Gaming; non-gaming shows the same shape at a larger scale, with its $4-7 million tier also rising considerably at 29%.
61%
Gaming video’s top 2% of creatives now capture 61% of all spend — up from 53%
Across every vertical and format, the top 2% of video creatives already account for roughly half of all spend or more, and by the top 10% that share climbs past 75% everywhere. Static images tell a different story: Gaming’s top 2% on that format actually eased from 51.5% to 46.6% over the same stretch.
74%
Action & Strategy games: Ad Networks lean 74% ‘Exciting’ on video, DSP flips to ‘Dramatic’ in image ads
‘Exciting’ creative tone leads video in this vertical across media types, but DSPs’ winning tone on images switches to ‘Dramatic’ (54.1% versus ‘Exciting’s’ 45.9%), the only case where the video and image winner for the same channel differ entirely. Action & Strategy’s ‘Key Selling Proposition’ is just as close a race, with ‘Benefit-oriented’ and ‘Feature-oriented’ splitting video spend 45% to 42%.

Social Proof Emphasis in Puzzle & Lifestyle games wins video by 69%

‘High’ emphasis takes up to 69% of video spend across channels, but on image DSPs’ share of ‘High’ drops to just 3.2%, with ‘Low’ emphasis taking over instead, up to 80%.

Features-centric content wins 43% of Finance’s image spend and 41% of video

Finance is the only vertical in the report where the same creative approach wins on both formats rather than flipping between them. That consistency suggests a more stable creative bet for Finance advertisers than most categories allow.

Shopping’s ‘Shop Now’ CTA rockets from 6% of video spend to 68% on image

On video, ‘Download’ narrowly leads a fragmented field at 29.2%, but on image ‘Shop Now’ becomes the dominant call to action, growing more than tenfold. It seems a fast, transactional CTA reads better as a single, immediate ask once footage is no longer competing for attention.
02 INTRO & GLOBAL TRENDS

Video and image are playing different games

Last year, this report described AI-driven creative production as a relentless numbers game: whoever generated the most variations, fastest, had the best odds of finding a winner. That premise still holds. If anything, the game has accelerated: production hasn’t slowed, and the advertisers already spending the most keep pulling further ahead.

Average creative counts at the top spend tier climbed 46% in a single year, while counts of advertisers with lower spend barely moved. Scale is compounding, not distributing, and the gap between well-funded and modestly funded creative teams keeps widening.

The gap between formats is not subtle either. Across gaming and non-gaming verticals alike, the attribute that drives the strongest video performance frequently reverses on static image ads: what reads as urgency in motion reads as noise in a frame, and the selling point that closes on one format falls flat on the other.

The two formats are also moving in opposite directions. Video’s top creatives are pulling further away from the pack, their share of spend climbing year over year, while image spend is spreading more evenly, the top creatives’ share easing over the same stretch.

That’s the real story underneath the numbers game: the volume advantage only compounds if it’s the right volume for the format it’s running in. The findings that follow — based on granular analysis of 1.2 million creative variations — show exactly where video and image diverge, and the one category where they still align.

How to read the ‘share of ad spend’ data

The data in this report shows how ad spend is split between creative types with multiple components. Since media platforms direct more budget to creatives that work, the share of spend is the best indication of creative performance.

The analysis compares video and static image ads, covering theme, hook, CTA, user generated content (UGC), gameplay, player motivation, and more — all identified by AppsFlyer’s AI-driven creative optimization engine, which automatically dissects numerous components of a creative to enable granular measurement for effective optimization.

Data sample *
1.2M
creative variations analyzed with AppsFlyer’s AI-driven creative optimization engine **
140
distinct creative elements ***
1.4K
apps

* All results are based on fully anonymous and aggregated data. To ensure statistical validity, we follow strict volume thresholds and methodologies and only present data when these conditions are met.

** A creative variation is set when there is the same visual in the same orientation. Any change whether localization, color, word, with or without a logo etc gets its own creative ID. Creatives are deduped when there is the same visual in the same type of orientation (e.g. portrait variations will be deduped assuming no other change).

*** A creative element is a single, specific value tagged within a creative attribute, for example ‘Exciting’ within Creative Tone, or ‘Get Started’ within CTA Type

Average creative variations per app by UA spend *

Cumulative share of spend by % of creative variations

03 Top trends – GAMING

Share of ad spend in Puzzle & Lifestyle games by media type

Puzzle & Lifestyle creative examples

Share of ad spend in Action & Strategy by media type

Action & Strategy creative examples

Share of ad spend in Casino by media type

Casino creative examples

Share of ad spend in Sports & Racing by media type

Sports & Racing creative examples

04 Top Trends – NON-GAMING

Share of ad spend in Finance by media type

Finance creative examples

Share of ad spend in Shopping by media type

Shopping creative examples

Share of ad spend in Media & Entertainment by media type

Media & Entertainment creative examples

Share of ad spend in Software by media type

Software creative examples

Share of ad spend in Food & Drink by media type

Food & Drink creative examples

Share of ad spend in Travel & Tourism by media type

Travel & Tourism creative examples

05 KEY TAKEAWAYS
Skip numbered cards section
Use AI tools to close the testing gap

Average creative counts at the $7M+ spend tier climbed 46% this year while smaller advertisers barely moved, and AI-assisted production and analysis is largely what's driving that gap. Advertisers without top-tier budgets should explore AI-assisted variation tools to keep testing volume competitive, rather than treating testing pace as fixed to spend size.

Treat video and image as separate bets

In nearly every vertical this report covers, the creative attribute that wins on video is not the one that wins on a still image. Evaluate the two formats separately rather than assuming a single winning direction transfers, and budget testing time for each independently.

Concentrate bets differently by format

Video's top 2% of creatives now capture 61% of spend, up from 53% a year ago, while image's top 2% share eased from 51.5% to 46.6% over the same stretch. As such, doubling down on analysis is key to finding success, especially in video.

Measure granularly for effective optimization

A single attribute can significantly swing spend share, which is why measuring performance down to the individual component, not just the overall ad, matters more than ever. As creative volume climbs, more data means more noise, not more clarity, and the full picture only emerges once you see which components combine into a winner.

Evaluate both surface and psychological drivers

Attribute-level analysis throughout this report shows that both surface elements, like CTA wording and visual style, and deeper psychological drivers, like emotional triggers and player motivation, can each swing performance significantly.

The videos and static image ads in this report were generated with the assistance of AI tools solely for illustrative purposes. They do not accurately depict any real person, place, or event but is an original artwork produced by artificial intelligence based on descriptive prompts. Any resemblance to actual individuals, living or deceased, or specific locations is purely coincidental. The content does not imply any endorsement, association, or connection with any natural persons or entities unless expressly stated therein.

Ready to start making good data driven choices?