The 3-second hook: anatomy of attention in 2026
Most ads die before the third second. Six archetypes, frame-by-frame. Scrub the timeline, then audit your own with the pre-ship checklist.
Why three seconds?
Imagine flicking through a deck of cards as fast as you can. Each card is an ad. Most of them you flick past in under two seconds - you barely see them. A few make you stop.
The ones that make you stop have something happening in the very first moment that breaks the flick. A surprise. A face looking at you. A problem you have. A result you want.
That moment - the first three seconds - decides if the ad gets watched. Everything after that only happens if the open earns it.
In one line: the hook isn't part of the ad. It's the entire reason the rest of the ad gets seen at all.
median time before scroll
of drop-off happens before 3s
hold rate gap, top vs bottom hooks
hook variants per concept (top accounts)
What's happening in those three seconds.
Three discrete viewer decisions, each running on a different mental clock. If the open doesn't earn each one in turn, the scroll resumes.
Is this for me?
Pre-attentive processing. Color, motion, face presence. Decided before conscious awareness.
Tactic: Pattern interrupt OR identity cue (face, logo, environment).
Is it interesting?
First conscious appraisal. Caption read, VO heard, scene framed. Curiosity gap opens or closes.
Tactic: Land the hook line - claim, problem-name, or POV cue. No throat-clearing.
Should I keep watching?
Promise-of-payoff check. Viewer commits or scrolls based on whether the next 10s feels worth it.
Tactic: Show the proof / demo / transformation will land. Tease, don't deliver yet.
Pick an archetype. Scrub the open.
The first 6 seconds, frame by frame. Click an archetype tab, then drag the timeline to see what should happen at each beat.
Pattern interrupt
S-tierFirst frame breaks visual expectation. Color, motion, or scene change in <100ms.
Hard-cut to high-contrast, unexpected image
Works when
When the feed context is predictable (e.g. another talking-head ad). Highest hold rate at second 1.
Fails when
Becomes pattern itself if every brand opens this way. Burns fast in saturated categories.
Anti-patterns
Five mistakes that kill hooks
The same five failures show up in low-performing libraries across categories.
Hook variation isn't a manual job anymore.
When you save an ad to Shuttergen, the analysis pipeline tags its hook archetype, identifies the frame-1 visual decision, and decomposes the open into the structural elements that make it work. From there, the remix engine generates 8-12 hook variants on the same proven scaffolding - different archetype, same hold-rate intent - so you can test six S-tier opens for the production cost of one.
That's the variation surface area Andromeda actually rewards: the hook structure changes, the product stays consistent, the audience stays unmistakably yours.
The checklist
Nine-step pre-ship audit
audit pass rate
Sources
What we read to build this
Stop shipping ads that die at second 2.
Save inspiration with the Chrome extension, let Shuttergen tag hook archetypes automatically, and remix into S-tier opens that actually earn the watch.
Get started free