The first three seconds decide everything. A viewer who keeps watching past the opening has implicitly agreed to give you the next thirty seconds; a viewer who swipes was never really yours. I used to treat hooks as an afterthought — write the video, then slap an opening on it. Reversing that order, and using AI as a hook-writing partner, changed my retention more than any editing trick ever has. This is the full system: hook formulas, open loops, and pattern interrupts, all built with AI assistance.
Why Hooks Deserve Their Own Writing Process
Most creators write a script linearly and hope the opening works. That is backwards. The hook is the highest-leverage sentence in your entire video — it deserves more writing iterations than any other line. I now write hooks last in one sense and first in another: the video concept comes first, then I generate and refine ten or more hook options with AI before committing to the body script, because the winning hook often reshapes what the video should emphasize.
AI is genuinely useful here for one specific reason: volume. A human writer falls in love with their first decent hook. An AI will happily produce twenty variants without ego, and among twenty variants there is almost always one that is sharper than your first instinct. The skill is not in writing one perfect hook — it is in generating options and recognizing the winner. For the broader script foundation this builds on, see my guide to going from idea to shoot-ready script with AI.
The Core Hook Formulas That Actually Work
Forget “hacks” — the hooks that consistently retain viewers fall into a small set of psychological patterns. Learn the patterns and you can generate infinite variations.
The Curiosity Gap
Promise valuable information while withholding the key piece: “The editing trick behind every viral transition takes eleven seconds to learn.” The viewer keeps watching to close the gap. The critical rule: the payoff must genuinely deliver, and it must arrive before patience runs out — ideally within the first third of the video. A curiosity gap with no payoff is just clickbait, and audiences punish it with the swipe.
The Contrarian Claim
Challenge a widely held belief: “Stop adding background music to your videos.” The disagreement reflex keeps people watching to see if you can defend the claim. This formula works best when you actually believe the contrarian take and can argue it — manufactured contrarianism collapses the moment a knowledgeable viewer shows up.
The Negative Avoidance
People are wired to avoid loss more than to pursue gain: “Three export settings silently destroying your video quality.” Loss-framed hooks consistently outperform gain-framed ones in testing across niches, because the fear of making a mistake is a stronger motivator than the hope of improvement.
The Direct Callout
Speak to a specific viewer: “If you edit videos on your phone, this is for you.” Being specific feels personal, and personal holds attention. The paradox is that narrower targeting often increases total retention — the right viewers watch longer, which is what the algorithm actually rewards.
The Proof-First Open
Show the result before explaining it: open with the finished effect, the before-and-after, the transformation. “Watch this clip go from flat to cinematic in one click.” Proof-first works because it replaces a promise with evidence — the viewer has already seen that the payoff is real. This pairs naturally with strong first-two-seconds hooks designed for the swipe decision.
Using AI to Generate Hook Options at Scale
Here is the practical workflow. Once your video concept is locked, prompt your AI writing tool with the concept plus a constraint: “Write 15 opening lines for a 45-second video about [topic]. Each must be under 12 words. Vary the approach: curiosity, contrarian, loss-avoidance, direct callout, proof-first.” The constraints matter — without them you get generic, interchangeable lines.
Then do the human pass. Read all fifteen out loud. Kill anything that sounds like every other video in your niche. Shortlist three. For each finalist, ask the AI for five variations that tighten the wording — “make this more specific,” “add a number,” “make it more urgent.” Specificity is the sharpener: “a video tool” becomes “this free editing app,” “soon” becomes “in eleven seconds.” The final hook should be the most specific version that is still true.
One more AI trick: ask for hooks your competitors would write, then deliberately avoid that territory. If every video in your niche opens with a question, open with a statement. Differentiation at the hook level is free, and it compounds across every video you publish.
Open Loops: Retention Beyond the Hook
A hook wins the first three seconds. Open loops win the next thirty. An open loop is an unresolved promise planted early and paid off later: “I will show you the exact setting at the end, but first you need to understand why the default fails.” The viewer’s brain dislikes unresolved threads — that mild tension keeps them watching through sections they might otherwise skip.
The professional structure is nested loops: a big loop spanning the whole video (“the setting I promised”), medium loops per section (“but that is not even the best part”), and micro-loops between sentences. You do not need many — two or three well-placed loops per minute of content is plenty. More than that feels manipulative; fewer and the middle sags.
Write your loops into the script deliberately rather than improvising them. I mark them in brackets during drafting — [LOOP: the free alternative] — so I can verify every loop gets closed before the video ends. An unclosed loop is worse than no loop: it leaves the viewer with the specific feeling of being tricked. Close every loop you open, on time, with the real payoff.
Pattern Interrupts: Resetting Attention Mid-Video
Attention decays. Even engaged viewers start drifting around the fifteen-to-twenty-second mark, and the script needs planned interruptions to snap them back. A pattern interrupt is any deliberate break in the established rhythm: a sudden question to the viewer, a visual change, a shift in vocal energy, an unexpected statement.
Script at least one interrupt every fifteen seconds of short-form content. The simplest interrupt is a direct question — “But does it actually work?” — which re-engages the viewer’s brain by demanding an internal answer. Visual interrupts work too: a zoom punch-in, a b-roll cut, an on-screen text pop. The key word is planned. Random interrupts feel chaotic; rhythmic ones feel dynamic.
AI helps here as an editor, not just a writer. Paste your finished script and ask: “Flag every 15-second stretch where the energy or topic stays flat, and suggest one pattern interrupt for each.” It is remarkably good at spotting the dead zones you have gone blind to, because you wrote them. This kind of structural review is also where planning a tight 15-second script teaches economy — every second has to justify itself.
Structuring the Full Retention Script
Putting it together, a retention-optimized short-form script has a clear architecture. Seconds 0-3: the hook, one sharp line. Seconds 3-10: context and the first open loop — just enough setup to orient the viewer, then a promise of what is coming. Seconds 10-30: the content in tight beats, each beat ending with a micro-loop or interrupt that pulls into the next. Final seconds: close the big loop with the real payoff, then the call to action.
Notice what is missing: throat-clearing. No “hey guys welcome back to my channel,” no “in today’s video I am going to.” The hook is the first sound the viewer hears. Every word before the value starts is a word the algorithm counts against you. When you build a shooting script and shot list from this structure, mark the hook and loop payoffs as your hero shots — they deserve the best visuals.
The Retention Read-Aloud Test
Before finalizing any script, read it aloud with a stopwatch running and mark where you feel your own attention dip. Those dips are your viewers’ swipe points. Add an interrupt, tighten the beat, or cut the section. If you cannot stay engaged reading your own script at a desk, no editing trick will save it on a phone screen. For longer narratives, the principles scale up — my breakdown of multi-shot 60-second storytelling shows how hooks and loops work across extended structures.
What AI Cannot Do For Your Hooks
An honest limitation: AI can generate options, tighten wording, and spot dead zones, but it cannot tell you what your specific audience finds irresistible. That knowledge comes from watching your own analytics — which hooks held viewers, which ones bled in the first seconds — and feeding those patterns back into your prompts. The creators winning with AI-assisted hooks are not using better tools; they are running a tighter feedback loop between their retention data and their writing process. Build that loop, and the hooks keep getting sharper every week. Start this week: generate ten hooks for your next video, pick the sharpest one, and compare its retention against your average — that single comparison teaches more than any formula ever will.