Thinking · Newsletter
AppLovin creative: the complete guide for apps, games and commercial brands (2026)
What AppLovin rewards and why Meta habits fail there: the published numbers on creative volume, length, captions and offers, how to produce 182 creatives from six concepts, why the landing page wastes most of it, and how to prove the channel adds demand.
TL;DR. AppLovin is a creative-supply channel, not a targeting channel. AppLovin's own guidance tells advertisers to upload as many 9:16 videos as possible and not to pause creatives; its June 2026 data put the average top campaign at 182 active creatives, and full libraries beat hand-curated sets by 7 to 23 percent. The videos that spend most run longer than 30 seconds, 95 percent carry captions, 85 percent use voiceover and about 40 percent carry an offer. Brands arriving with ten repurposed Meta ads hit a ceiling that looks like a channel problem and is a production problem. This guide covers what the channel rewards, how to produce the volume without producing 182 films, why the landing page is where most of the volume is wasted, and how to tell new demand from re-counted demand.
Who this is for
Anyone running or about to run paid acquisition on AppLovin for an app, a game or a commercial brand, and anyone whose Meta team has been asked to "also do AppLovin". It is written from the production side, because that is the side that decides the outcome.
What is different about AppLovin inventory?
Three structural facts, all from AppLovin's own published material, explain almost everything about creative on the channel.
The ad arrives in a high-attention gap, not a feed. A fullscreen video appears roughly once every two to five minutes of gameplay. Spend splits nearly evenly between two placements: interstitials, shown at natural breaks, where the viewer must watch at least five seconds before skipping; and rewarded placements, which the viewer opts into and which play through to the end, up to 60 seconds. AppLovin's CEO described it as "a high attention environment" and said social creative instincts "do not transfer cleanly".
The audience is in a game, not in a shopping mindset. Most are playing with the sound off. That is why captions are not optional here: AppLovin found that 95 percent of the top spending videos it reviewed had prominent captions.
The optimiser spreads spend across many creatives instead of consolidating it into one. AppLovin's support documentation is blunt: "You do not need to pause creatives. AppLovin's optimizer automatically reduces spend on creatives that don't perform. If you manually pause creatives, this may affect the creative's opportunity to select the best ad for each impression." A decade of Meta habit says kill the losers. Here the instruction is the opposite.
Sources: AppLovin, "Designing creatives for a full attention environment" (Adam Foroughi, 20 Feb 2026); AppLovin Support Center, "Video best practices".
How many creatives do you need on AppLovin?
More than you think, and the number is published.
In June 2026 AppLovin released performance data comparing curated and uncurated creative sets. Campaigns that used every available creative outperformed manually curated sets by 7 to 23 percent. Those campaigns averaged 182 active creatives, with a range of 65 to 340. The curated control groups ran between 5 and 10.
| Curated control | Full library | |
|---|---|---|
| Active creatives | 5 to 10 | 182 average, 65 to 340 |
| Relative performance | baseline | +7% to +23% |
AppLovin's own framing of why: "Many ads play a crucial supporting role in driving conversions, even if they aren't the last clicked AppLovin ad that gets attribution credit." An ad that looks mediocre on an aggregate report is often the only thing working in one pocket of a very large, heterogeneous inventory pool. Pruning it on attributed ROAS removes a piece of machinery that was earning its place.
Is 182 a target? No. It is the average of what worked in one dataset, and the range starts at 65. The directional reading is what matters: the working number is an order of magnitude above what most brands bring, and starting at twelve is what produces the conclusion "this channel doesn't work for us".
What does a top-performing AppLovin video look like?
AppLovin analysed its top 1,000 web-advertiser videos by spend. The recurring patterns:
| Pattern | Share of top-spending videos | Why it works here |
|---|---|---|
| Longer than 30 seconds | the recommended focus | Interstitials guarantee 5 seconds, rewarded plays to 60; short ads leave attention unused |
| Prominent captions | 95% | Sound is off for most players |
| Voiceover, first-person narrative | 85% | Longer storytelling that explains brand, product and value |
| A promotion or time-limited offer | about 40% | 10 to 40 percent off codes, free shipping, free item, "sale ending soon" |
On length, AppLovin's advice to teams without long-form assets is practical: stitch several UGC clips together, and showing more than one creator raises the odds the ad resonates with a wider audience. In the June 2026 dataset the top performers averaged 37.4 seconds.
Specs, for completeness: 9:16 portrait, MP4 or MOV, up to 60 seconds, up to 1 GB; end cards 9:16 at 320×480 up to 768×1024; playables a single HTML file under 5 MB with no external requests and no autoplay audio.
Sources: AppLovin Support Center, "Video best practices" and "Creative specs & guidelines"; AppLovin creative-set data, June 2026, as reported by Common Thread Co.
AppLovin vs Meta: what actually changes for creative?
Most comparisons argue about audience size and attribution. The decision that determines whether AppLovin works is a production decision.
| Meta | AppLovin | |
|---|---|---|
| Where the ad runs | Feed and stories, skippable | Fullscreen in-app, 5 second minimum or unskippable up to 60 s |
| Who sees it | People scrolling, often already in-market | People playing, mostly cold, sound off |
| What the algorithm does | Consolidates spend into a small winning set | Distributes across a large set; "do not pause creatives" |
| Working creative count | 5 to 15 is normal | 182 average in the June 2026 data |
| Video length | Short, front-loaded | Longer than 30 s recommended; top performers averaged 37.4 s |
| Repurposed assets | Travel well between placements | Consistently stall: wrong length, wrong placement, wrong register |
| The constraint when scaling | Budget and audience saturation | Creative supply |
Read the last row twice. On Meta, to scale you add budget. On AppLovin, to scale you add creative, and most brands discover this after deciding the channel does not work.
Why repurposed Meta assets stall, in order of damage: length (built to compress into three seconds, asked to sustain thirty); placement (built for a scroll that does not exist here, solving the wrong problem); register (a feed ad talks to a shopper, this viewer is mid-game).
How do you produce 182 creatives without producing 182 films?
This is where most brands stall, and it is not budget. It is production method.
182 creatives is not 182 shoots. It is somewhere between six and fifteen concepts, each built as a system that yields twelve to thirty finished variants without breaking the idea. Before anything is shot, write down what is allowed to vary:
- The opening. Five openings against one body is five assets from one shoot.
- The performer. The same script from three people is three assets, and it tells you something about your audience the first two did not.
- The problem frame. The same product solving a different stated problem.
- Length. A 15, a 30 and a 45 second cut of the same material, because the channel rewards the long one and the short one still earns its place in the set.
- The offer. With and without a code, since about 40 percent of top videos carry one.
- The end card. AppLovin added built-in end-card generation to Ads Manager in June 2026, so this axis got cheaper.
Six concepts, each yielding thirty variants, is 180 assets. That is arithmetic, and it is a production planning problem rather than a creative one. The question to ask any production partner is how many finished assets one shoot day yields. The answer tells you whether they think in films or in systems.
This is the operating habit mobile gaming user acquisition built over a decade: one title can carry tens of thousands of creatives a year, and the teams that win treat creative as supply. Prose, a hair-care brand, adopted exactly that habit with us: concepting stopped producing finished films and started producing concept systems defined by what can vary, so one approved idea became a family of assets rather than one asset.
Why does most AppLovin creative volume get wasted on the landing page?
Producing 182 specific creatives that all point at the same product page.
Volume buys specificity. Each ad is free to speak to one motivation rather than the average of all of them. Then every one of them lands on a destination built for the average, and the specificity you paid for evaporates on arrival. We read brand libraries for a living and this is close to universal: the ads are specific, the pages are generic, and the gap between what the ad promised and what the page opens with is invisible in any report that stops at the click.
HealthyCell is the clearest case we have. Before producing anything we read the existing library and every destination it pointed at: a library speaking to several distinct motivations, and a handful of pages speaking to none of them in particular. We built the ad and the page from one brief, as a matched unit, and tested them together. The point is not that landing pages matter; it is that the ad and the page are one decision, and splitting them across two teams guarantees they drift.
If you are about to scale creative volume, the cheapest thing you can do first is list every destination your current ads point at and count how many are product pages.
Why did my AppLovin account plateau?
Four causes, in the order we find them. Teams usually investigate them in reverse.
- Creative supply. Fewer than thirty active creatives and supply is the ceiling; nothing in the media settings lifts it.
- Repurposed assets at their limit. Built for a different length, placement and viewer. The plateau looks like a channel problem.
- Over-curation. Pausing on attributed ROAS, against AppLovin's explicit guidance.
- The destination. Every specific ad landing on the same generic page.
Three of the four are creative and destination problems, which is why we start there. The diagnostic: count active creatives; count how many were built for this channel; check what has been paused and why; list every destination and count the product pages.
Is AppLovin new demand or the same customers re-counted?
The right question, and it is measurable rather than arguable. Run a bake-off: your existing repurposed creative against creative built natively for the channel, at the same time, on the same channel, under the same conditions. If the native set merely reaches the same people, it performs like the baseline. If it clearly beats it, the channel is reaching people your existing creative was not. With HealthyCell, three native concepts against the repurposed Meta baseline outperformed on both interaction and purchase. Neither platform's self-reported ROAS should be taken at face value; incrementality tests and holdouts give a number you can spend against.
What to do this quarter
- Count your active creatives on AppLovin. Under 30, supply is your constraint.
- Stop curating down to winners. AppLovin says so and its data says so.
- Convert the next brief from a film into a concept system; write down what varies.
- Add captions to everything and let the long cut run.
- List every destination the ads point at before producing anything new.
- Run the bake-off so the incrementality question is answered, not argued.
Questions people ask
How many creatives do I need to start on AppLovin? Enough that the optimiser has something to distribute across. The published range of working campaigns starts at 65; AppLovin's guidance is to upload as many 9:16 videos as possible on a regular basis.
Should I pause underperforming creatives on AppLovin? No. AppLovin's documentation says the optimiser reduces spend on weak creatives automatically and that manual pausing can hurt the campaign's ability to pick the best ad per impression.
How long should an AppLovin video ad be? Longer than 30 seconds for most of the set. Interstitials guarantee five seconds and rewarded placements play up to 60, so short ads leave attention unused. Top performers in the June 2026 data averaged 37.4 seconds.
Do captions matter on AppLovin? They are close to mandatory. 95 percent of the top spending videos AppLovin reviewed had prominent captions, because most players have sound off.
Can I run my Meta ads on AppLovin? You can, and it is the usual starting point. Expect a ceiling from length, placement and register, and expect it to look like a channel problem.
What does producing this volume cost? It depends entirely on whether the supplier produces films or systems, which is why the same volume quotes wildly differently. Ask how many finished assets one shoot day yields.
What is Flixr's role in this? We came out of mobile gaming user acquisition, where this volume is normal, and we run the same system for apps, games and commercial brands: a creative lead inside your ad account, a read of your category in Flixr Studio, concept systems built to vary, and a weekly loop of brief, make, launch, read, log, judged on your numbers.
Sources: AppLovin Support Center, "Video best practices" and "Creative specs & guidelines" (2026); Adam Foroughi, "Designing creatives for a full attention environment", AppLovin blog, 20 Feb 2026; AppLovin creative-set performance data, June 2026, as reported by Common Thread Co.; Flixr client work with Prose and HealthyCell.