Ad hooks — the first two seconds, by mechanic

A hook is a VISUAL device, not a clever line. The 17 mechanics below are the ones that keep showing up in the longest-running ads across Meta, TikTok and YouTube. Use them by name when you brief or write an ad.

The mechanics

1. Direct callout — the creator leans into the lens and names the exact viewer in the first second ("if you can't fall asleep, watch this").
2. Open mid-problem — drop straight into the painful, relatable moment before any setup; the pain is the hook.
3. Before → after snap — an instant cut from the bad state to the good state inside ~2 seconds.
4. Object into frame — the product slides, rolls or is set down INTO frame with a confident thunk.
5. Big reaction — a genuine gasp, jaw-drop or "wait, WHAT" opens the clip; the emotion stops the thumb.
6. Satisfying macro — an oddly-satisfying extreme close-up: a pour, a peel, a snap, a texture.
7. Pattern interrupt — one unexpected, slightly absurd visual in the very first frame.
8. Provocative question — a bold, specific on-screen question or claim the target cannot scroll past.
9. Shock stat — a surprising number fills the screen as the cold open.
10. POV — first person; the viewer IS the character ("POV: you finally…").
11. Whip / snap-zoom — a fast whip-pan or snap-zoom that rushes into the subject.
12. Unexpected place — the product somewhere surprising, played completely straight.
13. Countdown to reveal — a quick 3-2-1 build that pays off on an immediate reveal.
14. List tease — promise a numbered payoff ("3 reasons I'll never go back to…") to open a curiosity loop.
15. Mishap → deadpan — a soft mishap in the first beat, then the creator carries on unbothered.
16. Street interview — a handheld-mic vox-pop; a stranger's hot take, then the product, then the reaction.
17. Surreal-calm setting — the creator calmly reviews the product somewhere impossible (a volcano rim, a skyscraper edge).

Picking one

• Talking-head / UGC: 1, 2, 10, 15, 16 fit naturally; 6 and 4 work when the product has a satisfying physical action.
• Product-only (no person on screen): 4, 6, 12, 13, 3 — never 1 or 10, which need a face.
• On-screen text hooks (8, 9, 14) only when the format will actually render text cleanly; on a plain video model, text garbles — composite it in post or skip these.
• Apps and software: 3 (before/after of the screen), 8, 9, 14; show the real interface inside 3 seconds.
• Regulated or serious categories (health, finance): avoid 15 and 17; prefer 2, 8, 9 with a specific, true claim.

Writing the beat

Write the hook as ONE sentence of action a camera can film, then the line spoken or shown, then the cut. Example: "Object into frame: a hand sets the bottle down on the marble with a thunk, label to camera. Line: 'Ninety minutes of hot yoga and it did not move.' Cut to the product in use." Then check: is the product on screen by second three?

Testing

Make three versions of the same ad that differ ONLY in the hook mechanic, run them against each other, and keep the mechanic that wins — the body of the ad is rarely the reason one loses.

Hermoso plans and renders ads with these mechanics built in (https://hermoso.ai/ai-ugc-ads/), but the list above is the whole method.