Correction, 2026-10-01: this page defined hook rate as three-second video views over impressions without a source. Meta Ads Manager calls the metric 3-second video plays, and TikTok Ads Manager has no 3-second count at all (it reports 2-second and 6-second video views); both are now cited. The page also stated which hook types cold and warm audiences "respond to" and that a 48-hour read window "is common" as facts we have not measured (they are now advice), and told you to set the presenter's expression "in the prompt", but Hailuo H3 Max Lip Sync takes no text prompt: the face comes from the still.
Correction, 2026-09-30: the first version of this page gave hook-rate benchmarks for Meta and TikTok as verified figures (we can't source them), said UGC-style creative out-hooks studio ads by a meaningful margin, and reported a hook rate from our own tests we can't reproduce. The definitions and examples stay; the numbers are gone.
A UGC hook is the first one to three seconds of a user-generated-content-style ad, the opening line, shot, and expression that stop a viewer from scrolling past. It is the part of a UGC video to test hardest: if nobody stops, nothing after it gets seen.
Everything downstream of the hook, the offer, the product shot, the call to action, only matters if the hook earns the stop first. A weak hook means the rest of your creative is never seen, no matter how good it is. That is the case for treating the hook as a separate unit to test: generate a dozen hooks against one body and let the data decide, rather than writing one script and hoping. The hook is not the headline; it is a visual and audio event that makes stopping feel like the correct choice.
How a UGC hook works
A working hook does three jobs in three seconds. It targets one person by calling out a specific viewer with a specific problem, so the people it excludes are supposed to scroll. It opens a gap, a problem unresolved, a claim unproven, a question unanswered, and that tension is the reason to stop. And it matches the visual to the emotion, because a confident line delivered over a confused face can read as fake.
Hooks fall into a handful of repeatable types:
- Problem-agitate: open on the pain, plainly stated.
- Curiosity: open a loop the viewer has to stay to close.
- Social proof: borrow credibility from a number or an authority.
- Contrarian: take the position the audience does not expect.
- POV / story: drop the viewer into a first-person moment.
Hooks are measured by hook rate, also called thumbstop rate, commonly calculated as 3-second video plays divided by impressions. It isolates whether the opening did its one job. The inputs differ by platform. Meta Ads Manager reports 3-second video plays: plays of at least 3 seconds (or nearly the whole video, if it is shorter), counted once per impression with replays excluded. TikTok Ads Manager has no 3-second count; its nearest columns are 2-second and 6-second video views, so build a TikTok hook rate from one of those and don't compare it directly with a Meta figure.
Benchmarks you'll see quoted vary by platform, category, placement and audience, so the most useful baseline is your own: the median hook rate of your account's recent ads on that platform. A new hook is a winner when it clearly beats that median at similar spend. Track hook rate alongside hook-to-click rate: a hook can stop the scroll without earning the tap.
When you use a UGC hook
Every UGC ad needs one, but pick the hook type for the audience temperature. A sensible first guess: problem-agitate and curiosity hooks for cold audiences who don't know your brand, since they create tension without needing context; social-proof and contrarian hooks for warm and retargeting audiences, who already recognize the product and need a reason to reconsider. Let your own hook-rate numbers overrule that guess.
Treat hooks as a testing surface. Hold the body and CTA constant, generate 8 to 15 hooks, ship them as a set, read hook rate after a fixed window you choose before launch, cut the bottom half and move spend to the two or three that separate. Hooks are cheap to produce and hard to predict from a desk.
Examples
Problem-agitate hook (cold audience, skincare). Opening line, delivered with a tired, slightly frustrated expression in soft bathroom light: "My dark spots were getting worse, not better." It targets one person (someone with dark spots), opens a gap (why were they getting worse?), and the frustrated delivery matches the line. What to check: that the face agrees with the words; a neutral or smiling delivery on a pain line reads as fake.
Curiosity hook (retargeting audience, supplements). Opening line with a conspiratorial lean-in: "I found out why my supplements weren't working and it's kind of embarrassing." It opens a loop the viewer has to close; put the sheepish half-smile in the presenter still's prompt, since in the route below the still sets the face. The rule: the payoff in seconds four to eight must be at least as strong as the tension in the hook, or you train viewers to distrust your next ad. A curiosity hook with a weak answer can post a high hook rate and a poor hold rate (how many viewers stay past the hook), which is why you read both.
What ten AI hooks cost on 8frame
One talking-head route: a presenter still, a voice line, and lip sync.
| Step | Tool | Credits |
|---|---|---|
| Presenter still | Nano Banana 2, 1K | 11 |
| Hook line (under 1,000 characters) | ElevenLabs text to speech | 13 each |
| Lip-synced clip, 5 s at 768p | Hailuo H3 Max Lip Sync | 54 each |
| Ten hooks | 681, about $6.80 |
ElevenLabs text to speech bills 13 credits per started 1,000 characters, so any hook line costs 13. Hailuo H3 Max Lip Sync bills per second of output, and the output runs as long as the audio, with a 5-second minimum charge: 34 at 480p, 54 at 768p (the default), 108 at 1080p and 216 at 2K for 5 seconds. A one-to-three-second hook line therefore bills as 5 seconds.
Lip Sync takes no text prompt, so the expression comes from the still. Reuse one presenter still for every hook that shares an expression, so the only variable is the line, and make another still (11 credits at 1K) for hooks that need a different face, such as frustrated versus sheepish.
FAQ
How long should a UGC hook be?
One to three seconds. The hook is the opening line, shot and expression; everything after it is the body.
What is a good hook rate?
Compare against your own account's median rather than a quoted industry number; benchmarks vary too much by platform, category and placement to transfer. A hook that clearly beats your median at similar spend is worth scaling.
Is hook rate calculated the same way on Meta and TikTok?
No. Meta Ads Manager counts 3-second video plays; TikTok Ads Manager reports 2-second and 6-second video views instead. Compare each platform against its own median.
How many hooks should I test at once?
Eight to fifteen against one body and CTA is a practical set: enough to find a separation, few enough to read with modest spend.
Related concepts
UGC hook formulas for AI ads is the copy-paste companion to this definition: hook formulas grouped by the five types above, each with a filled-in example.
Hook-to-click rate extends hook rate from the scroll-stop to the click. Reading both is how you tell a real winner from an attention trap.
Sources
- Meta's 3-second video plays definition: Meta Business Help Center, About video ad metrics on Facebook and Instagram, checked 2026-10-01
- TikTok's 2-second and 6-second video views: TikTok Ads Manager help, Video play metrics, checked 2026-10-01
- 8frame prices (Nano Banana 2 at 1K, ElevenLabs text to speech per started 1,000 characters, Hailuo H3 Max Lip Sync per resolution with a 5-second minimum): 8frame back-end tool registry, credit cost functions, read 2026-10-01
- Hailuo H3 Max Lip Sync takes an image and an audio track, no text prompt: 8frame front-end tool registry (Hailuo H3 inputs), read 2026-10-01
Generate ten hooks against one body, read hook rate, scale the winners.
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.