On this page
- The steps
- Real mobile app video ads
- What AI should and should not make
- Record the demo before you write anything
- Write three hooks, keep everything else fixed
- Assemble the ads in Symphony Creative Studio
- What long-running app ads open with
- Disclosure, safe zones and what to avoid
- Judge ads by what users do next
- Prompts to copy
- Three-hook test plan
- Mistakes to avoid
- Questions
Real mobile app video ads to model yours on
3,317 video ads from 298 advertisers in the Meta Ad Library, read October 5, 2026. The median one ran 38 days, and an ad still paid for after months is the closest public sign that it works. Live ads first.
- Ran 1 day
- Ran 2 days
- Ran under a day
- Ran under a day
- Ran 1 day
- Ran under a day
- Ran 2 days
+3,305 more mobile app video ads
Hooks, landing pages and run length for each
- Ran 1 day
- Ran under a day
- Ran 4 days
- Ran under a day
Every creative, with its copy, format and run length
The openers that ran longest
ΰ€ͺΰ€Ήΰ€²ΰ€Ύ ΰ€ΰ₯ΰ€² @Rs 1/- π
InstaAstro Β· running 1,395 daysCuentos en 23 idiomas para ti!
Beelinguapp Β· running 756 daysClassic merge puzzle game
Gossip Harbor Β· running 748 daysJoin Local Groups & Make New Friends
Meetup Β· running 590 days
What AI should and should not make
Use AI for the parts of a TikTok app ad that are about words and assembly: scripts, hooks, captions, voiceover and cuts. Do not use it to invent what your app looks like. Viewers decide within seconds whether the app is real, and a made-up interface fails that test while misleading people about features.
The working split is simple. A real screen recording carries the demo. AI writes three openers, drafts captions, and can generate short background shots for the first two seconds. TikTok's Symphony Creative Studio can generate scripts and video drafts, add captions, and export to Ads Manager, so one tool covers most of the assembly.
Keep every ad to 15 to 20 seconds. One app, one task, one result, one call to action. If you need a longer walkthrough, make it a separate ad instead of stretching the first one.
Record the demo before you write anything
Choose a task with a clear before and after. For a meal-planning app, that is picking meals and getting a grocery list. For a fitness app, it is starting a workout and seeing the finished summary. Record from the first tap to the result, with no logo intro and no tour of every feature.
Record on a real phone in portrait, with notifications off. Use sample data so no real person's name, email or photo appears on screen. Slow down your taps a little so a viewer can follow each one. You can speed the footage up in editing, but you cannot rescue a rushed recording.
Capture the result screen as a separate clean take of two or three seconds. That frame is your payoff, and you will reuse it at the end of all three versions.
- One task, not a feature tour
- Sample data only
- Portrait, notifications off
- A separate clean result take
Want mobile app video ads like these for your product?
The mobile app video ads that kept running, read from the ad libraries. Every change waits for your approval.
Write three hooks, keep everything else fixed
The first line is the only thing you should vary. Ask your AI writing tool for one problem-first hook, one result-first hook and one question hook, all built on the same real workflow. Tell it not to invent features, prices or testimonials, because it will otherwise add claims your app cannot back up.
Then lock the rest. The demo footage, the voiceover for the middle section and the call to action should be identical across versions. That way, any difference in results comes from the opening and not from a dozen small changes.
For a meal-planning app, the three openings might sound like this:
- Problem: "I'm tired of deciding what's for dinner at 6 p.m."
- Result: "Here's how I turn five dinners into one grocery list."
- Question: "What if your meal plan made your grocery list for you?"
Assemble the ads in Symphony Creative Studio
Bring your script and recording into Symphony Creative Studio, or into any editor you prefer. Place your real screen recording as the demo, add a voiceover, and turn on captions. Use the AI-generated draft as a starting point, not a final cut. Replace any generated app visuals with your own footage.
Use a tight structure. Seconds 0 to 3 are the hook, spoken and written on screen. Seconds 3 to 14 are the demo, one tap per beat. Seconds 14 to 18 show the result screen and one spoken call to action such as "Download it free and try it tonight." Only promise what the store listing supports.
Export vertically in 9:16. Check the TikTok ad specs for current file requirements, and preview each draft to confirm captions and key taps are not hidden behind TikTok's interface elements.
What long-running app ads open with
The Sprites ads library shows real ads from a category. In the mobile-app category, 3,317 video ads from 285 advertisers had a median run of 38 days. These are Meta ads, but the openers show what advertisers kept paying for. Meetup ran "Join Local Groups & Make New Friends" for 589 days with a Download button. Opal ran "A standing desk you can trust" for 558 days. Whatnot ran "Yup, you can buy new outfits without breaking the bank" for 386 days. Spotify ran "Give every step a soundtrack." for 331 days.
The pattern is one plain promise in the first line. Meetup names the outcome. Whatnot names the benefit and the objection together. Spotify ties the app to a moment, walking, that every viewer recognizes. None of them opens with a feature list or a brand intro.
Copy the structure, not the words. Write your hook as one outcome or one moment your user already has, then let the demo prove it. For more examples, see our guide to mobile app video ads.
Disclosure, safe zones and what to avoid
If you use fully AI-generated or significantly AI-modified visuals or audio, check that the required disclosure is present before you publish. An AI voiceover over a real screen recording is a different case from a fully generated scene, so review TikTok's current rules for each ad instead of assuming.
Keep generated footage away from anything that could mislead. Do not generate fake app screens, fake reviews, fake star ratings or realistic people who appear to endorse the app. Do not use a real person's likeness or another brand's logo in a generated shot.
If you want a human presenter, a creator-style ad usually works better than a synthetic one. See UGC ads for how to brief that, and keep AI for the hook b-roll, captions and cuts.
Judge ads by what users do next
Launch all three versions with the same audience and budget. If installs are your goal, compare cost per install. If what matters is a signup, trial or first completed task, set up tracking for that event and compare cost per completed action too.
A hook that earns cheap installs is not always the one that brings users who stay. A question hook may attract curious taps while a problem hook attracts people who finish onboarding. Read both numbers before you pick a winner.
Keep the winning hook and change one variable next: a new first frame, a different voiceover tone or a shorter demo. If you would rather not run this loop yourself, Director by Sprites is a full creative team for your startup. Every Monday you get finished, on-brand video ads, reviewed by a human. Learn more at Director.
Prompts to copy
Paste each one as written, then change the product, the setting and the light to match your own footage.
Script: three hook variants
Write three 15 to 20 second TikTok ads for [app name], a [type of app] for [target audience]. Use only this real workflow: [first tap, key action, result]. Give me one problem-first hook, one result-first hook and one question hook. Keep the demo and call to action the same in all three. For each, include voiceover, on-screen text and what to show. Do not invent features, prices or testimonials.
Hook: problem moment b-roll
Vertical 9:16, 3 seconds. A kitchen counter at dusk, a hand scrolling a generic phone with a blank screen, a half-written paper list beside it. Warm overhead light, slight handheld movement, shallow depth of field. No faces, no brand logos, no readable text. Leave the upper third of the frame clear for captions.
Hook: result-first reveal
Vertical 9:16, 3 seconds. Close-up of a hand holding a plain unbranded phone toward the camera, screen glowing white so a real screen recording can be placed over it later. Clean light background, soft daylight, slow push-in. No faces, no logos, no text on any surface. Centered composition with space above and below.
Hook: question with curiosity visual
Vertical 9:16, 3 seconds. A tidy desk seen from above, a generic phone face-up in the center, a coffee cup and notebook at the edges. Gentle camera drift to the left, bright natural light, calm color palette. No people, no brand marks, no legible writing. Keep the center open for a screen recording overlay.
Voiceover: calm direct delivery
Voiceover for a 16 second TikTok app ad. Tone: friendly, relaxed, like a friend showing you a phone shortcut. Pace: medium, short sentences with a brief pause after the hook. Read this script: [hook line]. [one-sentence demo narration, one tap per beat]. [result line]. [call to action]. No sales language, no exaggerated claims.
End card: call to action
Vertical 9:16, 2 seconds. A plain soft gradient background in [brand color], with a blank rounded rectangle in the center reserved for the app icon and a short empty text area below for the call to action. Subtle fade-in, no extra graphics, no logos, no people. Keep all key elements away from the bottom and right edges.
Three-hook test plan
| Variant | Opening style | Keep identical | Compare |
|---|---|---|---|
| A | Problem-first line | Demo, voiceover middle, CTA | Cost per install and cost per completed action |
| B | Result-first line | Demo, voiceover middle, CTA | Cost per install and cost per completed action |
| C | Question line | Demo, voiceover middle, CTA | Cost per install and cost per completed action |
| Next round | Winning hook, new first frame | Everything else from the winner | Change in cost per completed action |
Mistakes to avoid
- Letting AI generate fake app screens instead of using a real screen recording.
- Changing the hook, demo and call to action all at once, so you cannot tell what worked.
- Opening with a logo intro or a tour of every feature instead of one task.
- Placing captions or key taps where TikTok's interface will cover them.
- Skipping the AI disclosure check on fully generated or heavily modified visuals or audio.
- Picking a winner on cost per install alone without checking signups, trials or first completed tasks.
Rather have the videos made for you?
Director by Sprites is a full creative team for your startup. Every Monday you get finished, on-brand video ads, reviewed by a human. You never write a prompt.
Apply to Director by Sprites βQuestions
Can AI make a TikTok ad for my app without me recording anything?
It can draft scripts and generate background shots, but the demo should be a real screen recording. A generated interface can misstate what your app does. Record one task from first tap to result and let AI handle the words and assembly.
What is TikTok Symphony Creative Studio?
It is TikTok's AI creative tool for ads. It can generate scripts and video drafts, add captions and export to Ads Manager. Treat its drafts as a starting point and put your own screen recording in the demo section.
What aspect ratio should a TikTok app ad be?
Export vertically in 9:16. Keep captions and key taps clear of TikTok's interface elements, and check the current ad specs before you upload.
How long should a TikTok app install ad be?
Aim for 15 to 20 seconds. Spend about three seconds on the hook, most of the rest on one demo task, and the last few seconds on the result and call to action.
Do I need to label AI-generated TikTok ads?
If you use fully AI-generated or significantly AI-modified visuals or audio, check that the required disclosure is present. Review TikTok's current rules for each ad, since an AI voiceover over a real recording is a different case from a generated scene.
How do I know which AI-made ad is working?
Compare cost per install if installs are your goal. If a signup, trial or first completed task matters more, track that event and compare cost per completed action as well. The cheapest install is not always the user who stays.