Video Ad Hooks: Steal the Structure, Not the Script
The line expires. The shape underneath it does not.
Henry SedgwickProduct marketing
Search for video ad hooks and you get the same download every time: twenty five lines you can paste in front of any product. They work for about six weeks. Then every brand in the category is opening with the same sentence, the feed learns the shape of it, and the list stops earning anything.
The lines expire because the line was never the asset. What survives is the structure underneath: what sits in frame one, when the first cut lands, whether the claim is shown or stated, how long the loop stays open. Structures transfer between products and outlast every script built on them. So mine openings from ads that already have run time behind them, and copy the shape rather than the words.
A hook is a shape, not a sentence
The transferable part is frame one, cut timing and whether the claim is demonstrated or asserted. The words are the disposable layer.
Three seconds is the working unit
Meta counts a video play at three continuous seconds, and TikTok tells advertisers to introduce the proposition inside the same window.
Lines decay, patterns do not
A script everyone downloads saturates in weeks. Pattern interrupt and result-first have worked for as long as feeds have existed.
Longevity is your free quality signal
An ad still running after a month has survived a media buyer looking at it. Sort what you mine by that before you sort by taste.
Change one thing or learn nothing
If the whole creative changes between variants, no result can be attributed to the opening. Holding the rest fixed is the hard part, not the idea.
3 sec
Continuous play time Meta counts as a video play
15 sec
ThruPlay: completion below this length, fifteen unique seconds above it
$0
Cost of browsing competitor openings in Meta Ad Library and TikTok Creative Center
01
What a video ad hook actually has to do
A hook is not an attention grabber, it is a filter with three jobs to finish before the third second. Stop the thumb. Prove the video is about this particular viewer. Open something they want closed. Miss the first and none of the others run. Nail the first, miss the second, and you have bought a three second play from someone who was never buying.
The three jobs, and when each one is due
Most teams know all of this and still ship one opening per concept. Not from ignorance, from arithmetic: a second opening means the creator films again, or a sample ships to somebody in another country. The hook is the cheapest variable in the ad and, in practice, the most expensive thing in the account to actually change.
02
Six hook structures that keep working
Pattern interrupt
Frame one contains something the feed does not: an unexpected object, an odd angle, motion running the wrong way. Purely visual, so it survives the sound being off and needs no translation.
Claim inversion
Open by contradicting your own category promise, then spend the next ten seconds earning it back. Works precisely when the category is saturated with identical claims.
POV cold open
Drop the viewer into the middle of a scene with no setup, as if they arrived late. The missing first act is the open loop. This is the structure most genuinely native UGC runs on.
Problem in frame one
Show the pain happening instead of describing it. A demonstrated problem clears the relevance test faster than any sentence can, because nobody has to read anything.
Result first, then rewind
Lead on the outcome, then cut back to the start. The viewer stays to find out how, and the promise is carried in pictures rather than a claim you would have to substantiate.
Named callout
Open by naming the segment out loud so everyone else self-selects out. The cheapest structure to copy, which is exactly why it saturates first and decays fastest.
03
How to pull a hook structure off an ad that already works
- 1
Sort by survival before taste
Meta Ad Library shows the date an ad started running. Anything still live after thirty days has already survived somebody checking its numbers.
- 2
Watch the first three seconds muted
Five times with no sound, once with. Most of the hook is carried by the muted frame, and audio first makes you credit the script for work the picture did.
- 3
Write down what changes, not what is said
Frame one subject, camera distance, the timing of the first cut, when the product enters, whether a face appears at all. That list is the pattern.
- 4
Name the structure in four words
Result first, then rewind. Problem shown, no dialogue. If you cannot compress it into a phrase, you have captured a line rather than a mechanism.
- 5
Save the file and book the slot in the same sitting
Paused ads disappear from the library and hosted media URLs expire, so a bookmark decays into a dead reference. A pattern with no owner and no date is a screenshot.
04
Why a reference clip beats a longer hook brief
Here is where the week actually goes. You find an opening worth borrowing, so you describe it: punchy, handheld, real, product in shot by second two. Then the cut comes back and it is not the thing, and nobody did anything wrong. A brief encodes adjectives, and adjectives get re-interpreted by whoever holds the camera. What made the original work was timing and geometry, and neither has a word for it.
One opening, described versus carried
On video, a reference clip carries pacing, shot order, framing and the cues that read as native: handheld motion, natural light, the cut from talking head to product. Text can ask for UGC style and get a polished commercial back. A product reference does the other half, holding the label and silhouette identical between shots, killing the drift where the bottle quietly changes shape and the whole thing reads as fake. Fewer reshoots, openings that stay comparable inside one test, the tenth variant still matching the first.
Everybody can write a new hook. Almost nobody can shoot one this week, which is why the list of hooks is never the constraint.
05
Testing hooks so the result means something
A hook test only produces knowledge if the hook is the only thing that moved. Obvious, and broken constantly, because rebuilding the opening usually means rebuilding the whole ad, and then five things changed at once.
Two ways to run the same test
Swap the opening only
Rebuild the ad each round
That is the honest reason most brands run three openings a quarter instead of six a month. Not strategy, supply. The pattern library is free and takes an afternoon. Turning six named patterns into six real openings of your actual product, this week, with the packaging matching in every one, is the part that still costs a shoot.
06
Where AIMS fits
AIMS is built for that second half. It pulls competitor ads, organic posts and email into one searchable library, so the patterns you mine live somewhere you can produce from, then generates statics, product photography and video ads from those references with your product, palette and claims held fixed. One reference set feeds several openings on one concept, which is what makes a weekly hook cadence realistic without booking a creator per idea. Our AI UGC ads and AI product video ads pages walk through it.
07
FAQs
What is a good hook rate for video ads?
Hook rate is not an official Meta column. Most teams build it as a custom metric of three second video plays divided by impressions. Published benchmarks vary so widely by placement, country and objective that they are close to useless. Use your own trailing median as the baseline instead.
How long should a video ad hook be?
Three seconds is the practical unit. Meta only counts a video play once the video has run for three continuous seconds, and TikTok tells advertisers to introduce the proposition in the same window. Treat everything after second three as the body of the ad, not the hook.
Should I change the hook or make a new ad?
Change the hook first. It is the cheapest variable and usually the one carrying the difference, and holding the rest fixed is the only way the result tells you anything. Rebuild the whole ad once several openings have all failed.
Can I reuse a competitor hook?
Reuse the structure, never the material. Pattern interrupt, result-first and cold open are mechanisms nobody owns. The script, footage, music, logo and packaging belong to them, and copying those is a legal exposure as well as an obvious one to customers.
Why did my hook stop working?
Usually saturation rather than fatigue in the strict sense. A downloaded line spreads through a category in weeks and the audience learns to skip the shape of it. Structural patterns decay far more slowly than the wording layered on top.
Do hooks still matter on a fifteen second ad?
More, not less. Meta counts a ThruPlay as a completion when the video is fifteen seconds or shorter, so on a short ad the opening decides almost the entire outcome. There is no second act to recover in.