The Video Ad Structure That Sells: Hook, Pain, Solution, CTA

In short — A vertical direct-response video ad that converts follows four beats: hook (stop the scroll), pain or interest (make it their problem), solution (show the product doing the job), call to action (tell them exactly what to do). In a 15-second ad that's three shots of five seconds, and each shot carries 8 to 12 spoken words — no more, because that's what fits. Everything else is decoration. This guide covers the structure, the two frameworks worth using, and the production rules that keep the whole thing coherent.

Disclosure: Klaivo is our product — an AI ad generator built around exactly this structure. The word counts and shot rules below are the constraints we run in production, not theory.

Why structure beats production value

Creative quality is the single biggest driver of advertising ROI — roughly 56%, far more than targeting or reach (Nielsen). But "creative quality" for a direct-response ad doesn't mean cinematography. It means the ad does its job in the order a stranger's brain needs it.

A gorgeous ad with no hook is invisible. A plain ad with a sharp hook and a clear CTA sells. Structure is the part you control completely, whatever tool renders the shots.

The four beats

Beat 1 — Hook (0 to 2 seconds)

The hook exists to interrupt scrolling. Nothing else.

The window is brutally short, and platforms measure it differently: TikTok counts the hook at 2 seconds, Meta at 3 — which is why the same creative reports a higher hook rate on TikTok. TikTok for Business has reported that 63% of its highest click-through videos hook viewers within the first three seconds.

On Meta cold traffic, a hook rate of 25%+ is acceptable, 30%+ is good, and 35–45% is where scaled accounts operate, per benchmarks published by several performance agencies in 2026. Below 20%, the creative needs replacing, not optimizing.

Three hook patterns that work:

• The stated problem. "Stop wasting an hour on this."
• The contradiction. "Everyone gets this wrong."
• The visual interrupt. An unusual first frame — extreme close-up, unexpected motion — before a word is spoken.

What kills a hook: any form of greeting, any brand introduction, any slow build. "Hi guys, today I want to show you…" has already lost. Start mid-thought.

Beat 2 — Pain or interest (2 to 7 seconds)

Make it their problem, specifically. Vague benefit language ("premium quality, great value") converts nothing because it describes every product ever made.

Two paths here, and they're the two frameworks below: either agitate a pain the viewer already feels, or build desire by showing the outcome they want. Pick one. Doing both in five seconds does neither.

Beat 3 — Solution (7 to 12 seconds)

The product doing the job. For physical products this is the shot that decides the sale — the viewer needs to see the actual item, clearly enough to recognize what will arrive in the box.

This is where AI-generated ads most often fail: a model that invents a plausible-looking product instead of showing yours breaks trust at the exact moment you were asking for it. If your ad depends on the product being recognizable, this shot has to come from a real photo, not a text description.

Beat 4 — Call to action (12 to 15 seconds)

One action, stated plainly. "Tap to get yours." "Start today." Not three options, not a website URL nobody will type, not "link in bio" when the ad has a button.

Word counts: the constraint everyone gets wrong

The single most common mistake in scripted video ads is too many words for the time available.

A comfortable spoken pace is roughly two to three words per second. That gives you, per five-second shot:

Write more than that and one of two things happens: the voiceover rushes and sounds like a disclaimer, or the audio runs past the visual and the cut breaks. Eight to twelve words per shot feels uncomfortably short when you write it. It is correct.

AIDA or PAS: which framework to use

Two structures cover almost every direct-response ad.

AIDA — Attention, Interest, Desire, Action. Use it when the product delivers a clear positive outcome the viewer already wants. Over three shots:

1. Attention — pattern interrupt, no preamble
2. Interest and desire — the central benefit, and the viewer imagining having it
3. Action — a direct, result-oriented CTA

PAS — Problem, Agitate, Solution. Use it when the viewer has a specific, recognizable frustration. Over three shots:

1. Problem — name the irritant precisely, with real empathy
2. Agitate — the daily cost of not solving it
3. Solution — the product as the way out, then the CTA

How to pick: PAS is stronger when the pain is sharp and nameable (slow, painful, repeatedly failing). AIDA is stronger for aspirational or sensory products where nothing is really "wrong" — the viewer just wants the outcome. When in doubt, test one of each; they hook different people.

Production rules for AI-generated ads

Four rules that come from generating these at volume, and that don't appear in generic copywriting advice:

1. Keep one stable setting across all shots. A different room, lighting or background in every shot reads as a badly assembled cut, not a variety of angles. Describe a single, consistent environment for the whole ad — one minimalist studio, one warm kitchen — and let the camera move within it.
2. Write the visual prompt in English, whatever the ad's language. Video models are trained overwhelmingly on English descriptions and render more reliably from them. The spoken script follows your audience's language; the visual direction doesn't have to.
3. Silent motion in the visual prompt, voice on the audio track. Asking a model for a talking subject and then layering a voiceover produces mismatched lip movement. Specify closed lips and let the voiceover carry the words.
4. Lock the product before you polish the shot. Get the item rendering correctly first. A beautiful shot of the wrong product is worthless; a plain shot of the right one still sells.

The mistakes that cost the most

• Starting with the brand. Nobody scrolling knows or cares who you are yet. Earn the second beat first.
• One shot doing two jobs. A shot that hooks and explains does neither in five seconds.
• Burying the CTA. If it arrives after the viewer has left, it doesn't exist. Land it before the final second.
• Writing for a 30-second slot and cutting to 15. Structure to length from the start; trimming a long script produces an ad that sounds truncated.
• Testing one variant. Structure gets you a competent ad. Volume gets you a winner — the accounts that scale ship several variants a week and let the platform find the one that works.

The short version

Hook in under two seconds, with no greeting. One idea per shot. Eight to twelve spoken words per five seconds. Show the real product where the sale is decided. One clear action at the end. Then build a second variant, because the first one is a hypothesis.

Next: make a TikTok video ad without an editor applies this structure end to end, and what an AI video ad really costs covers the budget side of testing several variants.