Ad hooks: examples and patterns for AI video ads

How to read each hook pattern

Each pattern below has four parts: what it is, what the viewer sees in the first one to two seconds, a spoken line template, and the formats it suits. The templates are generic; replace the bracketed parts with the brand's product, problem and audience.

A hook has three layers that work together:

AI tools give you control over the visual layer that a phone shoot doesn't, so several patterns below lean on a strong first frame.

1. Question hook

The ad opens with a question the target viewer would answer yes to. It names the problem in their words.

First 1–2 seconds: a person close to the camera, looking straight at it, already mid-sentence. No intro, no logo.

Do you still [frustrating thing the audience does] every [time period]?
Why does nobody tell you [surprising fact about the problem]?

Suits: AI story-time ads, AI talking head ads.

Compliance: don't ask questions that assume a health condition or personal trait in the viewer. Ask about the situation, not the person.

2. Bold claim hook

The ad opens with the single strongest claim the brand can prove. It works only if the claim is true and specific.

First 1–2 seconds: the product in hand or the result on screen, with the claim as on-screen text.

This [product type] does [specific result] in [time it takes].
I replaced my [old solution] with this [product] and here is why.

Suits: AI product demo ads, AI talking head ads.

Compliance: use only claims the brand has evidence for and has approved. A generated person stating a result is still making that claim for the brand, and must not be presented as a real customer.

3. Confession hook

The speaker admits something slightly embarrassing that the viewer has also done. It reads as personal rather than as an ad.

First 1–2 seconds: a selfie-style shot, often in a car, bathroom or kitchen, the speaker looking a little sheepish.

Okay, I have to admit something about [habit or problem].
I spent [time period] doing [wrong approach] before I found [product].

Suits: AI story-time ads, AI get-ready UGC ads.

Compliance: the confession is a script, and the speaker is generated. Don't add text that says the person is a real customer or that the story happened.

4. Contrast or before-and-after hook

The ad opens on the problem state and cuts to the solved state, or shows both side by side.

First 1–2 seconds: a split screen or a hard cut between two states of the same thing, such as a messy room and a tidy one.

This was [problem state]. This is [result state] after [time] with [product].
Left is [old way]. Right is [product way].

Suits: AI product demo ads, AI green screen ads.

Compliance: show only a change the product actually delivers, at a realistic scale. Generated before-and-after images of bodies, skin or health results are high risk; check the platform's current policy for the product category before you make one.

5. Demonstration hook

The product does its main job in the first second, with no setup.

First 1–2 seconds: a close-up of the product working: pouring, cleaning, opening, fitting, switching on.

Watch what happens when I [use product on problem].
This is the fastest way to [task] I have found.

Suits: AI product demo ads, AI product placement ads.

Compliance: the generated product must match the real product's shape, size, colour and label. Check every frame for drift in the packaging text.

6. Curiosity gap hook

The opening line promises a piece of information and holds it back until later in the ad.

First 1–2 seconds: a speaker leaning in, or an object half-hidden in frame, with text that sets up the reveal.

There is one thing about [topic] that [group] never mention.
I tried [number] [product category] and only one did [result].

Suits: AI podcast ads, AI story-time ads.

Compliance: the ad must actually deliver what the hook promised. A hook that teases something the ad never shows reads as misleading.

7. Pattern interrupt hook

The first frame shows something the viewer does not expect to see in their feed. This is where AI video tools have the most room, because the image doesn't need to be filmable.

First 1–2 seconds: an object behaving oddly, a sudden scale change, a talking product, or a clay or 3D animated scene in a feed of filmed video.

[Product] here. We need to talk about your [problem].
Nobody expected [object] to [unexpected action] today.

Suits: AI mascot ads, AI claymation and craft ads, Pixar-style AI ads.

Compliance: stylised animation is clearly not real, which lowers the risk of misleading viewers. On TikTok, fully AI-generated video still needs disclosure.

8. Street question hook

A presenter with a microphone asks a passer-by a question tied to the product's problem.

First 1–2 seconds: a two-shot on a street, the mic already held out, the question as a caption.

Quick question: how much do you think you spend on [category] a month?
What is the one thing you would never [action related to problem]?

Suits: AI street interview ads.

Compliance: the people shown are generated, not members of the public. Don't present their answers as real opinions, and label the ad on TikTok. See AI-generated content labels on Meta and TikTok ads.

9. POV hook

The camera is the viewer's eyes, and the on-screen text sets up a situation they recognise.

First 1–2 seconds: a first-person shot, such as hands on a steering wheel or a view across a desk, with a "POV:" caption.

POV: you finally found a [product category] that [result].
POV: it is [time of day] and you just remembered [problem].

Suits: AI product placement ads, AI get-ready UGC ads.

Compliance: POV shots imitate phone footage. Keep them believable for the product, and disclose AI on TikTok.

10. List or number hook

The ad opens by promising a short list. Viewers stay to see every item.

First 1–2 seconds: a speaker holding up fingers, or the number as large on-screen text.

[Number] things I wish I knew before buying a [product category].
[Number] reasons [audience] are switching to [product type].

Suits: AI talking head ads, AI expert ads.

Compliance: "reasons people are switching" implies a trend; only use it if the brand can support it.

11. Objection-first hook

The ad opens with the viewer's main reason not to buy, said out loud, and then answers it.

First 1–2 seconds: a sceptical speaker, arms crossed or shaking their head, with the objection as text.

I thought [product] was too [objection] to be worth it.
Everyone says [common doubt about product type]. Here is what actually happens.

Suits: AI podcast ads, AI comedy skit ads, AI green screen ads.

Compliance: the answer to the objection is a claim. It needs the same evidence as a bold claim hook.

12. Expert framing hook

A speaker in a professional setting explains a problem the product addresses.

First 1–2 seconds: a desk, a lab coat, a workshop or a studio, with the speaker mid-explanation.

Here is what most people get wrong about [problem].
If you [common habit], stop and look at [specific detail].

Suits: AI expert ads, AI podcast ads.

Compliance: this is the highest-risk pattern for AI ads. TikTok does not allow AI content showing fake authoritative sources, even when labelled (TikTok Help Center). Don't give a generated person a real credential, a real name or a title such as doctor.

Hook patterns by format

Use this table to pick hooks for a format, or formats for a hook.

How to test hooks on the same angle

Test hooks by changing only the hook. Keep the angle, the body of the ad, the offer and the call to action the same, so that a difference in results comes from the opening.

1. Fix the angle. Start from one script with one angle, such as the main problem the product solves. Our UGC script template shows how to lay it out.
2. Write hooks from different patterns. For example a question, a demonstration and a confession, all leading into the same second line.
3. Make each hook as its own clip. With AI tools, generate the opening shots separately and cut each onto the same body.
4. Keep the person and setting the same unless the setting is part of the hook. Otherwise you are testing the person, not the hook.
5. Launch them together in the same campaign, with the same budget and audience.
6. Judge on the brand's own metric. Early watch time shows whether the hook holds attention; the brand's conversion metric shows whether it brings buyers.
7. Make more of the winner. Write new lines in the winning pattern, then make the winning hook in other formats, as in one winning script, made in every ad format.

On TikTok, if you duplicate a campaign to add new hook versions, the AI disclosure toggle resets, per TikTok's Ads Manager disclaimer guide. Turn it on again for every new ad.

Compliance checklist for any hook

Run these checks on the first two seconds before an ad goes live.

• The hook's promise is delivered later in the ad.
• Any claim in the hook is one the brand can support and has approved.
• No generated person is presented as a real customer, a real member of the public or a real expert.
• No public figure's face or voice, and no likeness of anyone under 18 (TikTok ad policy).
• The ad carries TikTok's AIGC label or your own clear disclaimer if any of it is AI-generated.
• On Meta, expect the AI info label on ad images made or significantly edited with AI (Meta help page).

At Anglr, the brief for each brand includes the angle and hooks, and every video is reviewed for QC and compliance before it goes live. If you make AI ads that perform and want to scale winning hooks across formats, apply as a creator. To see the formats these hooks open, browse the examples of AI ads.