A good hook shows your product or your promise in the first three seconds: a strong image, short on-screen text and a first line that also works without sound. Google, Meta and TikTok all recommend this [1][2][3]. You find the best hook by testing three to five openings, not by holding meetings.
In our studio for AI video production, we make several openings for one ad from the same product images, without a new shoot day [4]. The platforms also measure your ad from the first seconds: Meta counts 3-second views, TikTok 2-second and 6-second views, and a skippable YouTube ad can be skipped after 5 seconds [5][6][7].
Choose a hook type that suits your product
A hook does two things: it makes someone stop, and it says what the ad is about. A logo intro or a slow build-up does neither. Here are five types you can make from a few statics, worked out for a fictional stainless steel water bottle:
| Hook type | First 2 to 3 seconds | Pitfall |
|---|---|---|
| Show the problem | A lukewarm plastic bottle on a desk in the sun. Text: "Already lukewarm by 2 pm?" | The problem has to be real for your target audience |
| Show the result | Ice cubes still rattling when the bottle is opened in the evening | Only a result your product actually achieves [8][9] |
| Product in action | Macro: the cap twists open, ice cubes tap against the rim | With AI images: check shape, logo and colour |
| Price or offer | Packshot with "-20% until Sunday" in large letters | The offer must match your landing page exactly [8] |
| Direct address | The bottle in a bike's handlebar holder. Voice: "Do you cycle to work every day?" | Address one target audience, not everyone |
A question to the viewer, an unexpected image or social proof also work, as long as your ad answers the question, the image leads to the product and the reviews are real and verifiable [10].
Make three to five openings with AI
From second 3 onwards, the ad stays identical; only the opening changes. With AI, a new opening is a new shot, not a new shoot day.
- Choose different hook types, not five variants of the same line. TikTok recommends 3 to 5 creatives per ad group, clearly different from each other if you are testing [3].
- Generate only the opening shot of 2 to 3 seconds, from the static that fits, and let it flow seamlessly into the rest of your ad. The full step-by-step plan is in from product photo to video ad with AI.
- Put short text on screen. Google recommends supporting the message with both audio and on-screen text [1]; TikTok keeps it to 5 to 10 words per second [3]. Place the text within the platform's safe zone; see video formats by platform.
- Make your own subtitles in Dutch and French. Meta's automatic captions only work in English [11].
- Add sound, but make the hook work without sound too. According to Google, sound increases the conversions of Shorts ads by more than 20% [12].
Test one variable at a time
Compare your openings with the platform's testing tools, not by switching ads on and off yourself; according to Meta, that gives unreliable results [13].
- Meta A/B test: each version goes to part of your audience, and nobody sees both [13].
- TikTok Split Test: two equal groups, and a winner only at 90% confidence [14].
- Google video experiments: in Video Reach and Video View campaigns, with the creative as the only variable [15].
Meta built its ad system Andromeda for the growing number of creatives from generative AI [16]. The risk lies more on your side: too many variants with too little budget per variant, so that no test produces a conclusion. A winning hook also wears out; when it's time for new openings is covered in creative fatigue on Meta.
Measure the first seconds, but choose on cost per result
| Platform | Metric | What it measures |
|---|---|---|
| Meta | 3-second video plays | At least 3 seconds, replays not counted [5] |
| Meta | 3-second video plays rate per impressions | The share of impressions with a 3-second play, what marketers call the "hook rate" [17] |
| TikTok | 2-second and 6-second video views | At least 2 or 6 seconds, or an interaction within the first 6 [6] |
| YouTube | View (CPV) | Skippable in-stream: 30 seconds, the whole ad or an interaction; Shorts: 10 seconds or an interaction [7] |
Compare variants with each other, not with a benchmark: Meta publishes a definition of the hook rate, but no norm [17]. Figures like "we only have 8 seconds of attention" can't be traced; according to the BBC, that number didn't come from the report it was attributed to [18]. And a hook that holds viewers but sells nothing is not a winner.
What not to do
- Promise results you can't deliver. TikTok rejects or restricts ads that promise or exaggerate a product's effect, such as wrinkles disappearing in a before-and-after [8]. Misleading remains misleading, even with an AI label [9].
- Use an AI "customer" as social proof. According to the Belgian Advertising Code, testimonials must be genuine, verifiable and relevant [10]. Why is explained in AI UGC and AI avatars in ads.
- Conceal AI content. Disclose an AI opening on TikTok with the AIGC label or your own disclosure, otherwise the ad is rejected or restricted [8].
Want to see which openings your own product images produce? Send your statics for a free test ad: one video ad on your own brand, at no cost or obligation.
Frequently asked questions
What is a hook in a video ad?
The opening of your ad: the image, the movement, the on-screen text and the first line that make someone stop scrolling. A good hook also says right away what the ad is about.
How long can a hook last?
There is no official length. TikTok recommends putting the hook in the first 6 seconds and your proposition in the first 3, and a skippable YouTube ad can be skipped after 5 seconds. So design for the first 3 seconds.
How do I measure whether my hook works?
On Meta, divide the 3-second video plays by the impressions; on TikTok, look at the 2-second and 6-second video views. Compare variants with each other, because Meta and TikTok give no benchmark, and pick the winner on cost per result.
Can a hook exaggerate to stand out?
Standing out is fine, misleading is not. TikTok rejects or restricts ads that promise or exaggerate results, and a promise the product doesn't deliver remains misleading advertising, even with an AI label.
Is it true that we only have an 8-second attention span?
That figure can't be traced. According to the BBC, it didn't come from the Microsoft Canada report it was attributed to, but from a website with vague sources. Look at your own figures for the first seconds instead.
Sources
- About the ABCDs of effective video ads — Google Ads Help, accessed 27 Sep 2026
- About video ads — Meta Business Help Center, accessed 27 Sep 2026
- Creative best practices for performance ads — TikTok Ads Manager Help, 2025-06, accessed 27 Sep 2026
- AI video production (studio page: Google talk, variants per concept, test ad) — Izoard, accessed 27 Sep 2026
- About video ad metrics — Meta Business Help Center, accessed 27 Sep 2026
- Video play metrics — TikTok Ads Manager Help, 2025-11, accessed 27 Sep 2026
- About video ad formats — Google Ads Help, accessed 27 Sep 2026
- Misleading and false content (TikTok Advertising Policies) — TikTok, 2026-04, accessed 27 Sep 2026
- AI in advertising: transparency, disclosure and 12 clear principles (sensitive case: product performance) — Raad voor Reclame, 2026-08, accessed 27 Sep 2026
- De Belgische Reclamecode (Belgian Advertising Code), version 22 (Art. 20: testimonials and endorsements) — Raad voor Reclame, 2026-09, accessed 27 Sep 2026
- Add captions to your video ads — Meta Business Help Center, accessed 27 Sep 2026
- YouTube Shorts ads: Asset specs and best practices — Google Ads Help, accessed 27 Sep 2026
- About A/B testing — Meta Business Help Center, accessed 27 Sep 2026
- About Split Testing — TikTok Ads Manager Help, 2026-01, accessed 27 Sep 2026
- Create a video experiment — Google Ads Help, accessed 27 Sep 2026
- Meta Andromeda: Supercharging Advantage+ automation with the next-gen personalized ads retrieval engine — Engineering at Meta, 2 Dec 2024, accessed 27 Sep 2026
- 3-second video plays rate per impressions — Meta Business Help Center, accessed 27 Sep 2026
- Busting the attention span myth — BBC News, 10 Mar 2017, accessed 27 Sep 2026