You have about three seconds. Not because attention spans collapsed — because the platform shows the viewer a decision, and the decision takes about that long.
Most hooks fail for a boring reason: they spend the first second getting ready to say something. Here's a structure that starts already inside the point.
The formula
Specific situation + cost the viewer can feel + something they don't know yet
All three parts. Every time. If you drop the situation, it's abstract. If you drop the cost, there's no reason to care. If you drop the open loop, there's no reason to keep watching.
Why each part does work
The situation is a scene the viewer recognises, with a detail in it — a number, a habit, a time of day. Specificity is what makes it feel like it's about them and not about everyone.
The cost is what the situation is spending: hours, reach, credibility, money. Name it in the viewer's terms, not yours. "Reduced impressions" means nothing; "your last twelve posts did worse than the twelve before" means something.
The open loop is a genuine gap in their knowledge. Not a tease, not "wait till the end" — an actual missing piece they now want. A tease is a promise; a loop is a question.
Six worked examples
1. The volume trap
"You've made forty AI images this month. Three of them look like you. Here's why the other thirty-seven didn't."
Situation: a real number they could check. Cost: wasted effort. Loop: the cause they can't see from inside.
2. The misplaced effort
"Your hook is doing the work your thumbnail should be doing, and your thumbnail is doing nothing."
Situation: a division of labour they've never thought about. Cost: unspoken, which makes it sting harder.
3. The inverted effort
"Your best post took twenty minutes. Your worst took three days. That's not a coincidence, it's a diagnosis."
Time is the cost, and the inversion is the loop.
4. The near miss
"You rewrote the same prompt eleven times this week. The problem was one missing word."
Eleven is specific. "One missing word" is the loop — and it's small enough to feel findable.
5. The silent failure
"Your carousel gets saved and never read. Slide two is where it dies."
Saving is the flattering metric; reading is the real one. Naming the gap is the hook.
6. The subscription audit
"You're paying for four AI subscriptions and using the free tier of one of them."
Cost in money, situation in behaviour, loop in the obvious next question: which one?
The four habits that kill a hook
- Greeting. "Hey guys, welcome back." The viewer has not agreed to a relationship yet. You're spending their first second on your housekeeping.
- Throat-clearing. "So I've been thinking a lot about this lately…" This is a person getting ready to have an idea.
- Context before tension. Background is only interesting once something is at stake. Reverse the order and keep the background for the middle.
- Vagueness dressed as insight. "AI is changing everything." True, useless, and unfalsifiable. If nobody could disagree, nobody needs to watch.
Say the hook out loud, at normal speed, and imagine a stranger hearing only that sentence with no context. Would they ask a question? If they'd nod politely, rewrite it.
How to actually get one
Write twelve. Delete nine. The first three you write will be the obvious ones, and the obvious ones are what everyone else posts.
The prompt index has a "Hook stress test" that forces a rubric and a deletion round — it's a better use of a model than asking it for "10 viral hooks", which produces exactly the twelve you'd have written yourself.
A good hook on a bad idea buys you three seconds and then loses the viewer permanently. Fix the idea first. The formula amplifies whatever is underneath it.