Video ad hooks: 12 structures and how to test them

A hook is the first three to five seconds of a video ad, and it has three jobs: stop the scroll, tell the right viewer the ad is for them, and open a question the rest of the ad answers. Most hooks that work are built on a small number of structures. Below are twelve of them, each with a template, an example, what to put on screen, and when it fits. After that: how to test hooks and which numbers tell you one is working.

What a hook has to do in the first five seconds

Stop the scroll. This is mostly the picture. Movement in the first frame, a face close to the lens, a mess, a result, or a line of text large enough to read at a glance. A slow fade-in or a logo card gives the thumb time to move on.

Select the viewer. A hook that stops everyone is expensive. Naming the problem, the group or the situation lets the right person recognise themselves and lets everyone else leave before you pay for more of their time.

Open a question. The viewer should want the next ten seconds: what is the mistake, what did she switch to, how did the pan get clean. If the hook already says everything, there is no reason to keep watching.

A hook has two halves, what is said and what is shown, and they should do different work. If the line names a symptom, the picture shows it. Write both halves before you film or generate anything.

Twelve hook structures that recur

The examples are invented for ordinary products. Replace the brackets with the words your own customers use in reviews and support messages.

1. The symptom call-out

Template: "If your [who] does [symptom], [symptom] and [symptom], you need to hear this." Example: "If your dog scratches all night, licks his paws and smells yeasty, you need to hear this." On screen: the presenter to camera, with a quick insert of each symptom as it is named. In one real 35-second ad ReplicateAds measured, this hook runs 5.3 seconds over two shots with three photo inserts, one per symptom. Fits when: the buyer already sees the symptoms but has not connected them to a cause.

2. The mistake

Template: "The biggest mistake people make with [thing] is [common habit]." Example: "The biggest mistake with a cast-iron pan is washing it like every other pan." On screen: the mistake being made, in close-up, with a short overlay naming it. Fits when: your product replaces a habit people believe is correct.

3. The "I tried everything"

Template: "I tried [option], [option] and [option] for [problem]. Nothing worked until this." Example: "I tried foam rollers, stretching apps and a new chair for my lower back." On screen: the abandoned alternatives in a pile or a drawer, then the presenter holding the product. Fits when: the category is crowded and the buyer is tired of it. The speaker must be someone who really did try them.

4. The before and after, shown first

Template: [the result on screen] "This was [bad state] [time] ago." Example: a spotless oven door, then: "This was two years of baked-on grease twenty minutes ago." On screen: the after first, then a hard cut to the before. Fits when: the result is visible and it is your own real result, in your own footage.

5. The surprising claim

Template: "[Familiar thing] is [unexpected fact]." Example: "Your pillow is probably the oldest thing you sleep on." On screen: the familiar object, plain, with the claim as text. Fits when: you have a fact your page can back up. A surprise you cannot support is a liability, not a hook.

6. The direct address to a group

Template: "[Group], this one is for you." Example: "Runners over forty, this one is for your knees." On screen: someone who visibly belongs to the group, in the setting where the problem happens. Fits when: the audience is defined by who they are more than by a symptom, and broad targeting needs the creative to do the selecting.

7. The "stop doing X"

Template: "Stop [common action] if you want [result]." Example: "Stop scrubbing your face with a towel if you want the redness to calm down." On screen: the action, interrupted: a hand comes in and takes the towel away. Fits when: there is a clear action to replace. It is a command, so it wears out faster than the others. Rotate it.

8. The question

Template: "Do you [specific experience]?" Example: "Does your cat walk away from a full bowl?" On screen: exactly that experience, shown as the question is asked. Fits when: the honest answer for your buyer is an immediate yes. A question that takes thought loses the viewer.

9. The demonstration with no words

Template: [the product doing its job, close up, no speech for two or three seconds]. Example: a knife sharpener drawn once along a dull blade, then the blade through a tomato. On screen: the demonstration fills the frame, and a single line of text names what is happening. Fits when: the product's effect is visible and quick. It also works with the sound off.

10. The social proof opener

Template: "[Number] [people] have [done this], and here is why." Example: "Four thousand reviews, and most of them mention the same thing." On screen: your real review count or real review text, scrolled or held up on a phone. Fits when: the number is real and current. If it is not, pick another structure.

11. The price or comparison anchor

Template: "I was paying [amount] for [alternative]. This is [amount]." Example: "I was paying sixty a month at the salon for gel nails. This kit was thirty-five, once." On screen: the two prices side by side as text, or a receipt next to the product. Fits when: you are clearly cheaper than the thing you replace, and both prices are true on the day the ad runs.

12. The story start

Template: "[Time] ago, [something specific went wrong]." Example: "Last winter my hands cracked so badly I stopped going to the climbing gym." On screen: the place where it happened, or the presenter mid-activity, not posed. Fits when: the person telling it lived it. This is the structure for founders and real customers.

Test hooks by changing only the first five seconds

A hook test is only readable when the hook is the single difference between the ads.

1. Take one ad whose body you trust: the problem, the product, the proof and the call to action stay identical.
2. Make three to five versions that differ only in the first five seconds. Change the structure, not one word. A symptom call-out against a silent demonstration tells you more than two phrasings of the same question.
3. Change the picture with the line. A new sentence over the same opening shot is half a test.
4. Run them in the same ad set or ad group, with the same audience, budget and dates.
5. Read the early-view numbers first, then the sales numbers. Wait until each version has enough impressions that one good hour does not decide it.

The structure of a finished ad is described beat by beat in the anatomy of a winning UGC ad, which helps when you decide where the hook ends and the body starts.

How to tell a hook is working

Neither platform reports a metric called hook rate. It is a ratio advertisers compute from the metrics below. On Meta, open Ads Manager, then Columns, then Video engagement. On TikTok, they are the video play metrics in TikTok Ads Manager's reporting columns.

The definitions are from Meta's video ad metrics page and TikTok's video play metrics page.

The ratios most people use:

• Hook rate on Meta: 3-second video plays divided by impressions.
• Hook rate on TikTok: 2-second video views divided by impressions. The 6-second figure is a stricter second reading, but note that it also counts an engagement without six seconds of watching.
• Hold rate: ThruPlays divided by 3-second video plays on Meta, or views at 25% divided by 2-second views on TikTok. It tells you whether the people the hook stopped were the right people.

There is no official benchmark for any of these, and published averages vary with category, placement and audience. Use your own account: take the median hook rate of your recent video ads and judge each new hook against it. Read the two ratios together. A high hook rate with a weak hold rate means the hook stopped people it had nothing to say to, which usually means it promised something the body did not deliver. A modest hook rate with a strong hold rate and good cost per purchase is a hook that selects well. Keep it.

Borrow a competitor's hook structure, not its words

When a competitor has run the same ad for months, its hook has been tested with their money. What you can take from it is the structure: which of the twelve patterns it uses, how many seconds it lasts, how many shots it is cut into, and what kind of picture sits under each phrase. In the symptom example above, the transferable part is "three symptoms, one insert each, 5.3 seconds, then the turn". The dog, the creator and the sentence are theirs.

Write your version from your own page. List the symptoms your customers describe, pick the three that are most visible, and film or generate one picture for each. Do not carry over a claim because it sounded good in their ad: if your page does not support it, it does not go in yours. The whole process is in how to replicate a competitor's ad. ReplicateAds does this measurement for you, shot by shot, and rewrites the hook from what your page says before generating new footage.