7 hook patterns that stop the scroll in the first 3 seconds

1. Why the first three seconds are structural, not decorative

Short-form feeds do not give a video a warm-up lap. The viewer is already moving, and the only question the format asks is whether the next frame is more interesting than the swipe. That makes the hook a structural decision made in the edit, not a caption you slap on afterwards.

The scale involved is worth keeping in mind. YouTube has said Shorts average around 200 billion views a day — a figure Neal Mohan gave at Cannes Lions in June 2025 and repeated in his 2026 letter on the platform’s direction. Every one of those views is a swipe that did not happen. Your video is competing against a functionally infinite alternative.

Retention data from long-form gives a useful parallel. Retention Rabbit’s 2025 benchmark study, built on more than ten thousand videos, found that roughly 55% of viewers are gone by the sixty-second mark. Short-form compresses that curve into a fraction of the time. If you have not given someone a reason to stay by frame ninety, you have not lost them slowly — you never had them.

The practical consequence: you should be cutting the hook last, after the body is locked, and you should be willing to throw away the chronological beginning entirely.

2. The seven patterns

None of these are gimmicks. Each one is a different way of creating a small unresolved gap in the viewer’s head, which the rest of the video then closes.

Pattern 1 — The contradiction

Open by stating something the audience believes, then immediately undercut it. “Everyone tells you to post daily. We stopped, and the channel grew.” The tension is the gap between the two halves, and the viewer stays to find out which one is true.

Editing note: cut hard between the two clauses. No dissolve, no music swell. The abruptness does the work.

Pattern 2 — The in-progress open

Start mid-action, as if the viewer has walked in on something already happening. No greeting, no framing, no name. This is the single most reliable pattern we use, because it removes the setup that most creators instinctively film first.

Editing note: find the most kinetic two seconds in your footage and make it frame one. Then cut back to context after the hook has landed.

Pattern 3 — The stated cost

Lead with what something cost — money, time, a mistake. Specificity is what makes it work; a real number is far stronger than a vague claim. If you do not have a real number, do not invent one, use a different pattern.

Pattern 4 — The visual mismatch

The audio says one thing while the frame shows something that does not obviously match. The brain wants to reconcile the two, and reconciliation takes a few seconds. This one is purely an edit-room pattern; you cannot film it, you assemble it.

Pattern 5 — The direct callout

Name the exact person the video is for in the first line. “If you edit your own videos and it eats your weekend…” Narrow beats broad here. A callout that excludes 90% of the feed will hold the 10% far more tightly, and the algorithm rewards the depth, not the width.

Pattern 6 — The countdown promise

Tell the viewer precisely what they will have by the end and how long it takes. This is the least dramatic pattern on the list and the most reliable for educational content, because it trades tension for a clear contract.

Editing note: put the promise on screen as text at the same moment it is spoken. Redundancy is correct here.

Pattern 7 — The mid-sentence entry

Begin on the second half of a sentence. The missing first half creates a small backwards pull — the viewer keeps watching partly to work out what they missed. Use it sparingly; it stops working if every video in a feed does it.

3. The caption layer is part of the hook

A large share of short-form is watched with sound off or sound low, especially on a first impression. That means your hook has to work twice: once as audio, once as on-screen text.

• Put the hook text on screen within the first six frames, not after the first spoken word finishes.
• Keep it to one line. If the hook does not fit on one line at a readable size, the hook is too long.
• Never let captions sit in the bottom third — platform UI covers it. Middle or upper-middle, always.
• Change the caption on the beat of the cut, not between cuts. Movement should feel synchronised, not busy.

4. How long the hook should actually run

Since January 2025, when YouTube finished rolling out the creation tools for it, Shorts can run up to three minutes — the change was first announced in October 2024. Longer runtime does not mean a longer hook. It means the opposite: with more video to justify, the opening has to earn attention harder.

What we work to in practice:

• Frames 1–12: movement or a face already mid-expression. Nothing static.
• Frames 12–45: the hook line lands, spoken and on screen.
• Frames 45–90: the first payoff cut — a visual that proves the video will deliver what the hook promised.

If the first payoff has not arrived by three seconds, the hook is a preamble.

5. How to test hooks without guessing

Hook choice is not a taste question, and it is cheap to answer with data. Cut three versions of the same video where only the first three seconds differ, post them at least a few days apart, and compare how many viewers are still there at the ten-second mark rather than comparing total views. View count tells you what the feed did. Ten-second retention tells you what the hook did.

Do that four times and you will know which two patterns work for your specific audience — which is more useful than any general list, including this one.