Hook Them in the First Three Seconds

easy

Learn with your AI

Pick a depth. Each prompt opens in your AI pre-loaded with the lesson. Click a row to preview the prompt.

Or teach me the whole lesson:

Why this matters

Social ads (Reels, Shorts, TikTok) die in the first 2–3 seconds if they don't stop a scroll. Broadcast ads die in the first 5 seconds if the viewer reaches for the remote. The hook is not a 'trick' — it is the compressed promise of why the viewer should stay. AI tools let you produce 10 competing hooks in the time it used to take to shoot one, which means the craft of the hook is now the craft of the ad. The filmmaker who masters hook design out-ships the one who masters shot composition.

Demo

Hook archetypes that consistently perform: (1) Curiosity gap — 'I spent $0 on rent for a year, here's how' (2) Pattern interrupt — an unexpected visual in the first frame (3) Direct address — 'If you own an iPhone, this is for you' (4) Stakes front-loaded — 'This cost me $40,000 to learn' (5) Before/After tease — showing the after-state in frame 1, reveal the how. AI models handle visual pattern interrupts well; they handle direct-address scripts well; they handle curiosity-gap over-the-shoulder setups well. Write five hooks; shoot them all as 3-second clips; A/B test.

# Hook A/B test harness (5 variants per ad)
 
Product:      [X]
Proposition:  [one line]
 
Variant A — Curiosity gap
  Opening line:   "[ONE LINE that creates a question in the viewer's head]"
  Visual:         [what's in frame at 0:00–0:03]
 
Variant B — Pattern interrupt
  Opening line:   [bold visual claim]
  Visual:         [unusual image — wrong-scale, wrong-context, or unexpected subject]
 
Variant C — Direct address
  Opening line:   "If you [specific behaviour], watch this."
  Visual:         [shot of the specific person]
 
Variant D — Stakes front-loaded
  Opening line:   "Most people get this wrong. Here's what actually works."
  Visual:         [close-up; moderate motion]
 
Variant E — Before/After tease
  Opening line:   [show the end state first]
  Visual:         [frame 1 = destination; rest of ad = journey]
 
Testing plan:
  1. Generate all 5 at 3s each (cheap setting).
  2. Attach same middle + outro to each.
  3. Run as a Meta A/B test with matched budget.
  4. Watch-through rate at 3s is the signal. CTR is secondary.

Try it yourself

Scroll your own feed for 10 minutes. Note every ad where you stopped. For each, write the hook archetype. You will see the same 5 patterns repeat — because they work.

Write 5 hooks for the same product using the five archetypes. Shoot each as a 3-second AI pre-viz. Put them side by side and ask 5 friends which they'd watch next.

Pick the weakest hook and rewrite it until it beats the others. Understanding WHY a hook fails is more valuable than knowing which succeeds.

Audit a long-form ad you admire (60–90s). Does it front-load its hook in the first 5s, or bury it? The best long-form ads hook in the first 5s AND pay off at 60s.

Set a rule for your own work: no ad ships without 3 hook variants tested. Even informally — show to 5 people on Slack, measure who leans in.

Hook Them in the First Three Seconds

Why this matters

Demo

Try it yourself

References