Most teams treat hooks as a creative-team problem. They are a stakes problem, and they are a packaging problem. Almost everything that goes wrong with mobile ad hooks traces back to one of those two.
The stakes problem first. The four types of hooks that actually work in mobile UA — questions, bold stats, controversy, narrative — are not formulas. They are four different vehicles for the same payload: a specific stake the viewer recognizes in their own life. When the stake is concrete, any of the four can land. When the stake is generic, none of them do, and the ad reads as if AI wrote it because something close to AI usually did. This is why “Are you tired of…?” question hooks have not worked since 2018, why a stat-led opener like “73% of users churn after 30 days” gets ignored, and why “controversial” hooks that aren’t anchored in a real audience reaction come off as edgelord rather than insight. None of them are bad formats. They are formats running without their payload.
The packaging problem next, and this one is bigger than most teams admit. A hook is not a line of copy. A hook is the entire first three seconds. The spoken line, the visual cue, the on-screen text overlay, the audio register, the framing of the shot — all of it, together, as one package. A hook in isolation is not useful. A perfectly written spoken line filmed with the same camera setup, lighting, pacing, and text style as every other ad in the category disappears into the feed. Viewers do not skip it because the line was wrong; they skip it because the package was indistinguishable from a hundred ads they have already auto-rejected. The visual hook, the audio hook, and the text overlay are load-bearing parts of the hook, not decoration around the spoken line.
Both principles compound. A specific stake delivered in a generic visual package gets absorbed. A distinctive package without a real stake gets attention but no engagement. The hooks that scale carry both: a stake the audience recognizes from their own life, packaged in a first three seconds that looks, sounds, and reads visibly different from the surrounding feed.
The rest of this post applies those two principles to each of the four hook types, the failure modes for each, and a campaign structure that runs all four in parallel instead of betting on a single format.
Page Contents
What are the four types of ad hooks that actually work for mobile apps?
The four hook types are question hooks (a specific situation framed as a question), bold stat hooks (a number tied to a real customer insight), controversy hooks (a pattern interrupt anchored in genuine audience reaction), and narrative hooks (a three-second story with a concrete stake). Each is a delivery format for a stake. The format is interchangeable; the stake is not.
Use the table below to pick which format to lead with given the audience and the stake you are working with.
| Hook type | Lands when | Fails when | Best paired with |
|---|---|---|---|
| Question | The question names a specific situation the viewer is in (a stake, not a feeling) | The question is generic and sounds like 2018 Facebook ads (“Are you tired of…?”) | UGC delivery, talking-head openers, problem-aware audiences |
| Bold stat | The number was sourced from real customer language, reviews, or a credible external study | The number was invented to sound impressive (often AI-generated) | Analytical verticals, B2B/finance, comparison contexts |
| Controversy | The contrarian take is anchored in a real audience reaction the viewer recognizes | It is contrarian for its own sake — edgelord without a payload | Mature categories where viewers have heard the standard pitch a hundred times |
| Narrative | The three-second story has a concrete stake — a specific person, moment, and outcome | The story is archetypal and could be stock footage about anyone | Habit-forming apps, lifestyle apps, productivity apps |
What we see: The same axis decides every row — whether the hook carries a specific stake or a generic feeling. Format is downstream of that.
How do question hooks work and when should I use them?
Question hooks work when the question names a stake the viewer is already living. They fail when the question is a category cliche.
The failure mode is easy to spot. “Are you tired of…?” “Did you know that…?” “What if we told you…?” Any of these read as if they were generated by an AI working from a 2018 swipe file, because something close to that is what produces them. The viewer has seen this opener so many times it does not register as a question anymore. It is template noise.
The version that lands is specific in a way a category-level prompt cannot fake. Instead of “Are you tired of bad sleep?”, a working question hook for a sleep app is closer to “Why do you wake up at 3 a.m. and not at 6?” — because that is a stake the audience is in, on the night they are watching the ad. The question only works if it could not have been written from a generic brief.
The second move is delivery. Question hooks need a face and a tonal register that matches a real conversation. A polished voiceover reading the same question above stock footage produces the same template feel. Question hooks pair best with UGC, talking-head openers, or pattern-broken text-on-screen where the question feels overheard rather than scripted. The whole package — visual cue, framing, text — has to read as something a real person would say to another real person.
What makes a bold stat hook effective, and what numbers actually move the needle?
A bold stat hook lands when the number is grounded in real customer insight. It does not land when the number is manufactured to sound impressive, which is the default failure mode of AI-generated copy.
Manufactured numbers have a tell. They are round when they should be specific, specific when they should be approximate, and they almost always sit just above the threshold where a viewer would otherwise dismiss them. “73% of users feel…” “2.4x more likely to…” “Over 100 million people have…” A viewer does not consciously parse where these came from, but they do feel that nobody actually counted. The credibility load shifts from the number to the brand making the claim, which is exactly the wrong direction for a hook.
The opposite is a number with a paper trail. It comes from a customer review, a piece of public research, a published industry benchmark, or a thing the audience has independently heard before. The viewer does not need to verify the source — they need to feel that one exists. A stat hook that points outward at a real thing reads as evidence; a stat hook that points back at the brand reads as marketing.
This is why the cleanest source for stat hooks is customer language itself. The numbers people use when describing their own experience — “I tried it for three weeks before I saw any change,” “I lost two hours every morning to this” — are pre-verified by the audience because they came from the audience. The work is to mine reviews, comments, and survey responses for the numbers customers volunteer, then build the hook around the one with the most stake. (See how to find ad hooks in customer reviews using AI for the full mining process.)
How do controversy hooks drive engagement without damaging brand reputation?
Controversy hooks work because of pattern interrupt, not because of controversy. The line between the two is whether the contrarian take is anchored in something the audience already half-suspected.
The mistake is treating controversy as the goal. The goal is to interrupt the auto-skip behavior every viewer runs on every ad, and a contrarian claim is one of several ways to do that. The contrarian claim only earns the interrupt if it lands on a tension the audience is already living with. Stating “Your gym is making you fatter” without a real mechanism behind it is edgelord — the viewer skips faster, not slower, because they recognize the move. Stating something the audience has already complained about in their own reviews (“Most workout apps assume you have time you do not have”) is interrupt that pays off in the next two seconds.
The reliable way to source these angles is to read a stack of audience reactions to ads in your category — App Store reviews, comments under competitor creative, reaction posts. The friction the audience is already articulating is the controversy you can defensibly use. You are not inventing a new contrarian take; you are surfacing one that already exists in the market and pointing at it.
The second principle: a hook is never just the words. The visual cue, the on-screen text, the framing of the speaker, and the audio register all carry hook load. A “controversial” line read in the same camera setup, lighting, and pacing as every other ad in the category absorbs the controversy and flattens it. The pattern interrupt has to show up in the visual package too — a different shot type, a different on-screen text style, a different opening frame. Treat the entire first three seconds as the hook, not a tagline you wrote first and then filmed.
When should I use narrative or story-based hooks instead of direct claims?
Narrative hooks land when the three-second story carries a specific stake — a concrete person in a concrete moment with something concrete on the line. They fail when the story could be about anyone.
The failing version is archetypal. A vague person in a generic kitchen looks frustrated; a vague person in a generic gym smiles. The viewer’s brain recognizes the shape of an ad story without engaging with it. There is nothing to identify with because there is no specific stake — the person on screen could be replaced with any of a hundred other stock figures and the ad would not change.
The version that works front-loads specificity. A specific person (“the friend who never finishes the thing”), a specific moment (“the third week, when motivation runs out”), a specific stake (“you stop opening the app and quietly let it expire”). The viewer does not need the full story arc in three seconds — they need enough specificity that the story feels picked from a real life rather than assembled from a brief. Once the specificity hits, the rest of the ad has license to be more general; the hook has done its job.
Narrative hooks pair best with verticals where the product is a behavior change rather than a feature: habit apps, fitness apps, productivity apps, language apps, finance apps. In those categories the stake is the user’s relationship with their own follow-through, and a hook that names that stake outperforms a hook that names a feature, every time.
What we learn about running a hook portfolio
The four hook types are not a tier list. They are four different shapes for delivering a stake, and the campaigns that scale on Meta and TikTok run several of them in parallel rather than picking one and optimizing it to death. (Meta’s Andromeda update made creative diversity an explicit performance lever, not just a hygiene practice.) Three mistakes show up across teams that struggle with hook performance.
- Picking the format before the stake. “Let’s run a controversy hook this sprint” is the wrong starting point. The right starting point is the stake — the specific situation the audience is in — and the format choice falls out of which one delivers that stake most directly.
- Treating the hook as the words alone. A working hook is the entire first three seconds: the visual frame, the on-screen text, the speaker’s energy, the audio. Writing the line first and then filming it as a normal ad is what produces the “good copy, dead delivery” failure mode.
- Sourcing hooks from internal whiteboards instead of from the audience. Hooks pulled from team brainstorms tend to surface what the team finds clever. Hooks pulled from real customer language, reviews, and audience reactions surface what the audience already feels. The second source wins more often than the first.
Campaign structures that make sense
If you are running a hook portfolio rather than chasing a single winning hook, the structure below is a reasonable starting point. It assumes you are on Meta or TikTok with at least four creative concepts in rotation at any time.
- One concept per hook type, minimum. Run at least one question, one stat, one controversy, and one narrative hook concurrently. This gives you four distinct shots at the audience and lets the algorithm select.
- Source each hook from a different stake, not a different angle on the same stake. Four variations of “your sleep is broken” is one hook, not four. Four variations sourced from four different audience tensions (sleep, energy, focus, mood) is a portfolio.
- Ship the visual package, not the line. For each hook, brief the first three seconds end to end: shot type, on-screen text style, speaker register, audio cue. Different hooks should look visibly different in the feed.
- Refresh on hook fatigue, not concept fatigue. When CPI starts climbing, a fresh visual on the same hook does not unstick it. Replace the hook with a different stake, then film it.
- Read your own reviews and your competitors’ reviews weekly. The next hook is almost always already in the comment section. (For the full mining workflow, see how to find mobile ad hooks in customer reviews using AI.)
How do I test which hook type performs best for my specific app?
Run all four in parallel rather than sequencing them, and read the results at the hook level rather than the ad level. A clean test ships four creative concepts in the same campaign — one per hook type, each carrying a different stake, each filmed with its own visual treatment — and lets the platform allocate.
The metric to watch is not which ad wins. It is which hook type plus stake combination wins, because the next four creatives you ship should compound on the winning combination rather than the winning execution. (For the underlying campaign-structure rationale, see how to structure Meta campaigns for creative testing.)
Which hook types work best for different app categories?
The hook types are not category-locked, but the audience disposition in a category does shift the odds. Below is the directional read.
- Subscription apps (health, fitness, language, finance, productivity): narrative hooks anchored in a follow-through stake tend to outperform direct claims because the product is a behavior change.
- Gaming (casual, hyper-casual, mid-core): question and stat hooks tied to a specific game mechanic land harder than narrative; the stake is “can you do this thing” rather than “are you the kind of person who…”.
- Fintech and B2B: stat hooks dominate when the number is sourced from a credible external benchmark or customer language. Manufactured numbers are punished hardest in this audience.
- Social and dating: controversy and narrative hooks work because the audience tension is already public. Mining audience reactions is straightforward.
- Utility (camera, scanner, photo apps): question hooks tied to a specific use case (“the receipt you cannot read three months later”) outperform branded claims.
None of this is a mandate. It is a starting prior. The fastest way to falsify it is to run all four hook types against your audience and let the data overrule the prior.
Frequently Asked Questions
What is the most important factor in a mobile ad hook?
The presence of a specific stake the viewer recognizes from their own life. Hook type, format, and visual treatment are all downstream of whether the stake is concrete or generic. A stake-specific question hook outperforms a stake-generic stat hook every time.
Why do AI-generated hooks fail in mobile UA?
They tend to default to category-level language (‘Are you tired of…?’, ‘Did you know…?’) rather than situation-level language. The viewer has seen the template a hundred times and skips before the hook lands. The fix is to source hooks from real customer reviews and audience reactions, not from a generic brief.
Is a hook just a line of copy?
No. A hook is the entire first three seconds of the ad — the spoken line, the visual cue, the on-screen text overlay, the audio register, and the framing of the shot, working as a single package. A well-written line filmed with generic delivery and a stock visual treatment absorbs into the feed and is skipped just as fast as a poorly written line. The visual hook, the audio hook, and the text overlay are load-bearing, not decorative.
How many hook types should I run in a single campaign?
At least four — one of each hook type — with each hook delivering a different stake rather than four variants of the same one. Meta’s Andromeda update rewards this kind of portfolio diversity, and TikTok’s algorithm responds similarly.
Are controversy hooks risky for brand reputation?
Only when the controversy is invented for shock value. A controversy hook anchored in a real audience reaction (an actual reviewer complaint or a recognized industry tension) reads as insight rather than edgelord, and the brand effect is positive. The risk is in inventing a contrarian take with no audience grounding.
How often should I refresh hooks?
When CPI starts climbing, replace the hook with a different stake before refilming. A new visual on the same stake fatigues at the same rate as the original. Hook-level rotation outpaces ad-level rotation.
What is the fastest way to find a working hook for a new app?
Read 50 to 100 customer reviews of the closest comparable apps in your category and write down every concrete stake the reviewers volunteer. The hooks worth testing are almost always already in that pile. This works because customers have already done the audience research for you.
Do narrative hooks need to be acted out, or can they be voiceover-only?
Either works if the specificity is there. The format is less important than whether the three-second story names a specific person in a specific moment with something specific on the line. A voiceover that delivers all three with no visual story can outperform a fully shot narrative that delivers none of them.
Related Reading
- How to find mobile ad hooks in customer reviews using AI — the mining workflow that feeds every hook type above.
- Why creative diversity matters for ad account health — Andromeda’s view of why portfolios beat champions.
- How to structure Meta campaigns for creative testing — the campaign-structure side of running a hook portfolio.
- How to hire a creative strategist for mobile UA — the role that translates customer language into hook briefs.
- The Ultimate Guide to AI-Driven Creatives — where AI helps in the hook process and where it ruins it.
