Every creative team looks at what competitors are running. The difference between research and browsing is whether anything comes out of it besides a vague sense that everyone else's ads look better. Playables make this harder than it should be, because they are the format ad research tools handle worst: most public ad libraries were designed to archive images and video, and a playable is neither.
This is a working method: where to look, what those sources can and cannot tell you about an interactive ad, the signals that suggest an ad is actually performing, and what is worth taking from it.
The free libraries, and what each is for
Meta's Ad Library is the most complete free source. It is public, needs no account, and lists active ads across Facebook, Instagram, Messenger and the Audience Network for any advertiser, searchable by page. Each ad shows the date it started running, which, as below, is the most useful single field in it.
Google's Ads Transparency Center covers verified advertisers across Google's surfaces, including Search, YouTube and Display. You can filter by region, date and format, but the formats are text, image and video, and there is no spend, targeting or performance data.
TikTok offers two different things. The Creative Center's Top Ads is a curated gallery of high-performing ads by region, industry and objective: good for what is working in a category, poor for tracking one competitor, since a brand appears only when one of its ads ranks. The Commercial Content Library, built in response to the EU's Digital Services Act, is a searchable repository of ads shown in the EU, with the creative, run dates and some targeting and reach information.
What they cannot show you about a playable
There are two blind spots, and both matter.
First, the libraries archive a representation of the ad rather than the ad. For a playable that is usually a video capture or a still frame. You see the art, the hook and often the end card. You do not see how soon the first interaction is asked for, how hard the mechanic is, whether there is a fail state, or when the call to action appears, which is most of what makes a playable work.
Second, most playable spend runs in in-app inventory (Unity, AppLovin, ironSource, Mintegral and the rest) which the platforms' own transparency tools do not cover at all. Paid ad-intelligence services sample that inventory and capture playables from it, and they are worth the subscription for a team shipping playables every week. For everyone else there is a crude and effective alternative: install a handful of apps where your category advertises, play them with ads on, and screen-record every playable that serves. It is the only way to feel the interaction the way a viewer does.
Signals that an ad is working
None of the free tools show results. What they show is behaviour, and an advertiser's behaviour leaks their results.
- Longevity. An ad that has been running for weeks is very likely profitable, because nobody keeps paying for a loser. A start date is the closest thing to a performance number a library offers.
- Iteration. Several variants of one concept (same mechanic, different colours, different hooks) means the advertiser found something that works and is now optimising it. That is a stronger signal than any single ad.
- Spread. The same concept appearing across regions, placements or languages means it survived a test and was scaled.
- Recurrence. A concept that disappears and comes back months later is one the advertiser trusts.
A single sighting means nothing
A new ad you see once is a test, and most tests lose. Research built on one-off sightings copies the losers as often as the winners. Weight what has run for a long time and what has been iterated on, and ignore what you have seen only once.
What to extract: a teardown template
Write each ad down in the same structure. Patterns across twenty ads will appear that no single ad reveals.
- Hook: what is on screen, and moving, in the first second.
- Time to first interaction: how long before the viewer is asked to do anything.
- Mechanic: which of the common archetypes it is, and how many actions it takes to reach the resolution.
- Fail state: whether the viewer can lose, and what happens if they do.
- Resolution and call to action: when the button appears, what it says, whether it persists.
- End card: the offer, the app icon, any rating or social proof.
- Format: orientation, length, languages, and whether it pairs with a video.
What to take, and what to leave
Take the structure: pacing, time to first interaction, how long the loop runs, when the call to action appears and how it is phrased. These are properties of the format and of your category's audience, and they transfer between products.
Leave the art, the characters and anything recognisably theirs, which is a legal problem as well as a creative one. And leave any mechanic your product does not contain. Every platform's policy requires the ad to represent the app it promotes, so copying a competitor's mechanic into an ad for a game that does not have it is exactly the misleading creative those policies are written about. See the policy guide. Remember too that by the time you have seen it in a library, their audience has seen it as well.
Turn observations into tests
Research is useful only when it ends in a hypothesis small enough to test. Category leaders ask for the first tap within a second and we ask after three: that is a test. Their end cards carry a star rating and ours do not: another. Change one variable against your current control, run it long enough for the result to settle, and keep the teardown sheet as the record of why each test was run. The A/B testing guide covers the discipline.
A routine that keeps it useful
- Weekly, spend half an hour on five named competitors in Meta's Ad Library and Google's Transparency Center, noting new concepts and what is still running.
- Monthly, scan TikTok's Top Ads for your category and region, looking for structural patterns rather than individual ads.
- Quarterly, run a device session in apps your category advertises in and record the in-app playables that serve.
- Add every ad worth keeping to the teardown sheet, in the same structure every time.
- Review the sheet with the creative team, pick one or two hypotheses, and test them.
When a concept is worth testing, the fastest route is usually a template close to its mechanic, re-skinned with your own art, rather than a build from nothing.
Create playable end cards in minutes—no code required.
Open Playable Ads Maker