Nothing in an app ad is believable until the screen is on camera. Three script shapes, twelve hooks and the store listing handoff that keeps the install.
A mobile app UGC ad has one job the other categories do not: it must show the screen doing the thing, on a real phone, in the first two seconds. Everything else follows from that. Say what the app costs, show the one tap that solves the problem, and make the store screenshots match the frames you just showed.
Not a mockup, not a motion graphic of a phone floating in space. A screen recording with a thumb moving over it, with the viewer's own status bar habits visible: a low battery, a notification, an ordinary mess of apps. Polished demo data is the fastest way to look like a promo.
Ads that walk through five tabs sell nothing. Pick the single action the app exists for, and film it end to end at real speed. A viewer needs to see a problem go in and an answer come out. Feature tours belong on the store page, not in a fifteen-second ad.
The top comment under every app ad is some version of "it is a subscription". Get in front of it: say the monthly price out loud and put the paywall screen on camera for a second. An ad that hides the price buys installs that churn on the first open, which is the most expensive kind.
A cracked screen, one hand, a bus window, glare in direct sun. Those details do more for credibility than any script line, and they also prove the app is usable in the situation the buyer will actually be in. An app filmed only at a clean desk looks like it was never used outside one.
The store listing is your landing page, and the handoff is where installs leak. The screenshots a viewer meets first should show the same screen the ad showed, in the same order. Put the ad's promise in the opening lines of the description, before the text is cut off, and check how your listing renders on the devices you are buying on.
Beats with timecodes, then a written-out example. Swap the product name and the detail, keep the shape.
A utility app whose entire value is one action that is faster than the manual way.
Product: Kerbside, a parking sign reader (fictional)
[Screen recording, already scrolling. A cracked corner on the screen is visible.] "I got a ninety-dollar ticket because I misread a sign."
[Thumb opens a camera app, photographs a sign through a windscreen, squints at it.] "Two arrows, three times, and a day I did not understand."
[Opens Kerbside. One tap. Photo of the sign. A line appears: can park until 6pm, 2 hour limit.] "One photo. It reads the sign and gives me one sentence."
[Scrolls to the paywall screen, holds it.] "Three ninety-nine a month. That is the whole paywall, no surprises."
[Holds the phone up.] "Kerbside. Search Kerbside."
Spoken word count: 74. Runtime: about 25 seconds.
An app bought out of a specific frustration, where the buyer's own store search is the hook.
Product: Mileform, a mileage log for self-employed drivers (fictional)
[Screen recording of the App Store search field. Typing is visible.] "I typed 'mileage log no spreadsheet' at eleven at night in March."
[Scrolls results. Opens one, closes it. Opens another, closes it.] "That one wanted my bank. That one was ads."
[Opens Mileform. Taps once. A drive starts logging. Cut to the end of the drive, a figure appears.] "One tap when I get in. It stops when I park. That is it."
[Taps into settings, shows the export screen and the price line.] "It exports a CSV my accountant accepts. Four ninety-nine a month."
[Phone held up in a van.] "Search Mileform. One word."
Spoken word count: 79. Runtime: about 27 seconds.
A subscription app where retention is the pitch and the buyer expects to abandon it.
Product: Tidepool, a weekly budget app (fictional)
[Phone on a kitchen table, morning light, a hand steadying it.] "I have deleted six budgeting apps. This one is on day seven."
[Screen recording of the onboarding, thumb tapping skip twice.] "Eleven onboarding questions. I skipped nine of them and it still worked."
[Opens to a single screen with one number on it.] "This is the only screen I open. One number, what I have left this week."
[Scrolls to a cluttered settings page, pulls a face.] "The categories screen is a mess and I never touch it."
[Shows the subscription line.] "Two ninety-nine monthly. Cancel and it drops to read only, it does not lock my data."
Spoken word count: 87. Runtime: about 29 seconds.
The opening line carries the ad. These are written to be said out loud by a person holding a phone.
The wrong face kills a good script. Match the person to the buyer, not to your taste.
A student filming in a dorm with a cracked screen and a chaotic home screen. Thumbs fast, talks about what she typed into the store and which apps she deleted first. Credible because she looks like someone who installs and abandons apps weekly, which is who the ad is trying to convert.
Self-employed, filming from the driver's seat of a parked van with sun glare across the screen. One-handed, short sentences, no interest in features. Credible for utility apps, because he is visibly using the phone in the exact awkward situation the app claims to handle.
Holds the phone at arm's length at a kitchen table and reads the screen aloud as she taps. Not fluent with apps and says so. Credible for anything sold to people who bounce off complicated onboarding, because she narrates every place she nearly gave up.
In this category the proof is a recording, not a sentence. Match the promise to the footage before you write a word of voiceover.
| The promise | The recording that earns it | What fails |
|---|---|---|
| It is faster than doing it by hand | Both methods on the same phone, timed at real speed | "Saves you hours every week" |
| It works with no signal | Airplane mode toggled on screen, then the action still completing | "Works offline" over a stock shot |
| The subscription is fair | The paywall screen held for a full second while you say the price | A price in six-point text at the end |
| It is easy to start | The onboarding recorded unedited, including the parts you skipped | A polished three-frame animation |
| My data comes out | The export screen, and the file opening in another app | "Exports anywhere" |
If a claim has no recording, cut the claim rather than illustrating it with a graphic.
The click does not land on a page you control the layout of, and that is what makes this category different. The viewer arrives at a store listing with a title, an icon, a screenshot strip and a description that gets truncated.
Three things to align before the ad runs. The first screenshots should show the screen the ad ended on, so the viewer recognises where they are. The opening lines of the description should repeat the ad's promise in the ad's words, not in brand language. The price and what the subscription includes should be findable without scrolling to the bottom.
Then check the listing yourself on the devices and regions you are buying in, because the layout and the amount of visible text differ between them.
Partly, and this is the one category where the generated half and the real half divide cleanly. The screen recording has to be real. The face saying the hook does not.
On MakeViral's AI creator page a persona costs 2 credits on Nano Banana 2 or 1 credit on Z-Image, and a talking lip-synced video on Kling Avatar 720p costs 2 credits per second of speech, from 3 seconds to 15. An 8-second hook is 16 credits. Persona plus hook plus the 2-credit hook-and-demo stitch that joins it to your screen recording comes to 20 credits, about $1.43 on the $49.99 plan for 700 credits and about $1.33 on the $199.99 plan for 3,000.
Skip the product-in-hand still here. There is no physical product to hold, and a generated phone in a generated hand showing a generated interface is exactly the mockup this category punishes. MakeViral has no ad account integration, so the finished clip is a download you upload wherever you buy installs. See the AI avatar video generator.
App ads get judged on two things at once: what the ad says, and whether the app behind it matches. Both have to hold.
A creator you generate is not a real user, so the script is your pitch and not a review, and it must never be framed as one person's experience of your app. Rules for disclosing synthetic creators change, so check the current requirements on the platform you are buying the ad on before it goes live.
A B2B viewer sorts by job title in the first line and by migration pain in the fifth. Three script shapes, twelve hooks and the screens that answer both.
Read morePlaybookOne rep from the side beats a montage every time. Three script shapes, twelve hooks, and the footprint and noise tests that renters actually ask about.
Read morePlaybookThe delivery moment is this category's native opening. Three script shapes, twelve hooks, and the scale and color shots that stop returns before they start.
Read moreAll UGC ad playbooks · AI influencer generator · AI UGC ad generator
Pick a creator, paste the script, and get a talking video. Two credits per second on Kling Avatar, two credits for the creator portrait.
No watermark. A failed video gives its credits back. Secure checkout by Stripe.
