Video Hooks: The Complete Guide to Scroll-Stopping Openers

Short answer: a video hook is the opening one to three seconds of a video (the first visual, the first on-screen text, and the first words) designed to stop the scroll and earn the seconds that follow. Strong video hooks open on something already happening, make one specific promise in under eight words, and cut to proof fast. Nothing else in your video matters if the hook fails, because almost nobody sees anything else.

This is the pillar guide for our hooks series. It covers the mechanics end to end: why hooks work, the seven types that cover most short-form openers, why reaction hooks have become a default for app and product ads, and the testing loop that separates teams who guess from teams who know. If you just want lines to copy, jump to our 150+ hook ideas swipe file; the map below points to the deeper guide for each platform and hook type.

Video hooks guide map

Why the first two seconds decide everything

Feeds reward attention. TikTok, Instagram, and YouTube all report early retention and watch time, and all describe engagement signals as inputs to what they show next. A strong hook compounds twice: people keep watching, so the platform has a reason to show it to more people, who keep watching. A weak hook fails twice the same way.

The practical implication: the hook is usually the highest-leverage variable in the ad because it changes how many people ever reach the offer, the demo, and the call to action. It is also the cheapest part to change, which is why most testing programs start there.

The psychology: pattern interrupts and faces

Scrolling is a trance. Feeds train a rhythm — swipe, glance, swipe — and a hook works by breaking the pattern the brain has predicted. Three interrupts reliably beat the trance:

1. A reacting face. Faces are the one stimulus the brain is hardwired to process before conscious thought. An expressive reaction — shock, disbelief, delight — triggers an instinctive pause while the viewer reads the expression. Test reaction-led openings, including clearly represented AI UGC, against other formats.
2. An open loop. A claim or question that creates an information gap ('wait, this does *what*?'). The discomfort of an unclosed loop is mild but real, and watching is the only way to close it.
3. Motion and cut discontinuity. Starting mid-action, mid-sentence, or mid-fall signals 'you arrived late to something happening' — and arriving late is more compelling than being welcomed at the door.

The viewer doesn't decide to watch your video. They fail to decide to skip it — and then it's too late, they're curious.

The three layers of every hook

A hook isn't a sentence. It's three channels firing at once, and the strongest hooks make them agree:

• Visual layer — what's on screen at frame one. A reacting face, a surprising object, a result shot. In practice it does most of the work, because viewers register the image before they read a word.
• Text layer — the on-screen caption. Under 8 words, top third of the frame, native-looking font. It names the loop the visual opened ('Wait, this app does THAT?').
• Audio layer — the first spoken words or sound. On sound-on platforms this is a second hook for a second audience; write it so the video also works muted.

The 7 types of video hooks

1. The reaction hook

A 3–5 second clip of a filmed or AI-created character reacting — jaw-drop, skeptical squint, shocked laugh — followed by a hard cut to the product or demo. The face earns the pause, the caption frames the claim, the demo pays it off. More on why this one dominates below.

2. The bold claim

'This replaced a $500/month tool.' Big, specific, falsifiable — the demo must immediately start proving it or the viewer feels baited and leaves worse than they came.

3. The question

'Why is nobody using this?' Questions force an internal answer, and answering is engagement. Best when the question is one the viewer has privately asked themselves.

4. The negative / warning

'Stop doing this immediately.' Leans on loss aversion: people tend to react more strongly to avoiding a loss than to gaining something, which makes warnings worth testing on cold paid traffic. Pair the warning with a fix and you have a problem-solution hook.

5. The POV / story drop

'POV: you finally found the cheat code.' Drops the viewer into a scene as protagonist. Native to TikTok grammar and the most resistant to ad blindness — it doesn't resemble an ad at any frame.

6. The result-first

Show the outcome before the process: the 240k-view analytics screen, the finished room, the after photo. Works when your outcome is visually undeniable; tease it instead when it isn't.

7. The mid-action start

Begin in the middle — mid-sentence, mid-unboxing, mid-fail. No context, no introduction. The missing beginning *is* the open loop.

Why reaction hooks dominate app and product ads

For software and apps there's a structural problem: screens don't stop scrolls. A screen recording — however beautiful the product — reads as UI, and UI reads as ad. The reaction hook solves this by putting the emotional payoff first: a reaction that illustrates the feeling your message is about, without claiming the character experienced your product, then a hard cut into the demo while the viewer's curiosity is still open.

1. Seconds 0–3: reaction clip + bold caption. The face is the pattern interrupt; the caption opens the loop.
2. Seconds 3–8: hard cut to the demo. No fade, no logo. Show the single most 'wait, what?' capability first.
3. Seconds 8–20: one feature, kept moving. Every second must re-earn the next.
4. End: soft callout — name, 'link in bio', or let the result speak.

The economics are the quiet advantage. Filming custom reactions means creators, briefs, and per-clip fees; a licensed reaction-clip library gives you many interchangeable openers for roughly the price of one custom video, which turns hook testing from a production problem into a selection problem. That's the model behind RE/UGC's library of hyperrealistic AI UGC reaction clips. Agents can search and download clips too: see how to build UGC ads with Claude. For the full sourcing comparison, see the UGC ads playbook.

Platform notes: TikTok vs. Reels vs. Shorts

The hook testing loop

1. Fix everything except the hook. Same demo, same length, same CTA. One variable.
2. Ship five variations. Different hook types, not five wordings of the same idea — a disbelief open, a POV, a warning, a price shock, a result-first.
3. Run 48–72 hours. Long enough for the algorithm to settle, short enough to keep shipping.
4. Judge on hold rate first, CTR second, CPA last. Hold rate tells you the hook worked; CTR tells you the demo did; CPA tells you the offer did. Diagnose in that order.
5. **Iterate on the winner's *mechanism*, not its words.** If the price-shock hook won, test three more price angles — you've learned what your audience flinches at.

Product / feature: [one feature only]
Audience: [who, and what they already believe]
Fixed body: [demo file, length, CTA - identical across variations]
Hook A (disbelief): visual [ ] / text [ ]
Hook B (question): visual [ ] / text [ ]
Hook C (warning): visual [ ] / text [ ]
Hook D (price or result): visual [ ] / text [ ]
Hook E (POV): visual [ ] / text [ ]
Primary metric: 3-second hold vs. account average
Secondary: CTR, then cost per result
Decision date: [launch + 3 days]

For how many variations your budget can actually support, see how many ad variations to test.

Three hooks, broken down frame by frame

Example 1: the disbelief open (app ad)

Frame 1: a woman mid-gasp, hand moving to her mouth — motion already in progress. Text: 'Wait, it does this automatically?' Second 3: hard cut to the app auto-sorting an inbox. Why it holds: the face fires the pattern interrupt, the caption converts the emotion into a specific claim, and the demo starts paying it off before the viewer's skepticism catches up. The whole persuasion sequence happens in five seconds.

Example 2: the skeptic's open (testimonial)

Frame 1: creator with arms crossed, side-eye at the camera. Text: 'I really wanted this to fail.' Seconds 2–8: 'Everyone kept recommending it, which honestly made it worse…' Why it holds: the viewer's own skepticism is acknowledged before it can trigger a skip. Openers that *agree with the doubter* convert doubters — enthusiasm at frame one only converts the already-convinced.

Example 3: the result-first open (ecommerce)

Frame 1: an analytics screen showing a real view count from the brand's own post, shot over a shoulder on a phone, deliberately imperfect. Text: 'One hook did this.' Second 2: whip cut to the creator pointing at the exact opening clip. Why it holds: the outcome is undeniable and visual, and the open loop ('*which* hook?') is closed only by watching. Note what all three share: zero logos, zero warm-up, text under 8 words, and a face or a number in frame one.

The five mistakes that kill hooks

• Opening on a logo or brand sting. The most expensive frame of the video, spent announcing 'this is an ad.'
• Warming up. 'Hey guys, so today I wanted to talk about…' is four seconds of dead air. Start mid-thought.
• Caption over the chin or behind UI. If the platform's interface covers your text, the hook doesn't exist. Top third, always.
• Two ideas in one hook. A hook making two claims makes zero. One loop, opened cleanly.
• Bait without payoff. A shock hook the demo can't cash triggers instant exit — and the algorithm reads the exit as a quality verdict on the whole ad.