A repeatable hook-writing system

The first line decides whether anyone reads the second, so it deserves disproportionate effort. The mistake is treating hooks as a spark you wait for. Treat them as a system you run — generate many across proven patterns, sharpen the best — and you'll never stare at a blank first line again.

Strong hooks aren't random; they pull a few psychological levers. The curiosity gap opens a loop the reader needs closed. The contrarian take challenges a held belief. The specific number signals concrete value. The vulnerable admission ('I was wrong about X') earns attention through honesty. Once you know the levers, you can ask AI to generate across all of them and pick what fits.

The line you must not cross: a hook makes a promise, and the content must pay it off. Clickbait that over-promises gets the click but burns trust — the only durable asset in content. This is also why you write the hook last: you can only honestly promise what the piece actually delivers once it exists. Draft the content, then mine it for the most compelling true thing to lead with.

When a hook of yours performs, or one in the wild stops your scroll, save it. Over time you build a personal swipe file of proven structures you can hand to AI as examples: 'write hooks in the style of these'. Models imitate concrete examples far better than abstract instructions, so a swipe file quietly raises the ceiling on every hook you generate.

1. 1Draft the post first, so you know what it delivers.
2. 2Ask AI for 10 hooks across different patterns, under 12 words each.
3. 3Pick the two strongest and make sure the post pays them off.
4. 4Save the ones that performed as a swipe file.

Hook-writing prompts