UGC Ad Hooks for AI Video: 6 Formats, Prompts and a 3-Hook Test
Pick a format, type your opening line in the Hook field (the person on screen says it), and generate one clip per hook to A/B test openings from 8 credits each.
Open the UGC Ad StudioThis guide covers only the opening: hook formats, a prompt for each, and a clean three-hook test. For the full production checklist see the AI UGC video ads guide; for choosing a workflow, the AI video ad maker overview; for organic tactics, 10 TikTok growth hacks.
What the platforms actually say about hooks
What the platforms themselves publish (checked September 26, 2026):
- TikTok: "Introduce your content proposition in the first 3 seconds for better recall and awareness," and "prioritize your hook in the first 6 seconds." It suggests "suspense, surprise, or other emotions," a "DIY or not overly polished style," and 9:16 at 720p or higher (TikTok creative best practices).
- Meta: ads "should communicate clearly without sound, but get better or funnier with sound on." Meta reports that adding captions increased video view duration by an average of 12% and suggests delivering the message in 15 seconds or less (Meta for Business).
- Testing: TikTok lists "the initial 'hooks' (the opening 2-3 seconds) of your video ads" as a variable you can split test (TikTok split testing variables).
So the first line must work as on-screen text as well as speech, and the product or problem must be visible in the first shot.
How the hook works inside the UGC Ad Studio
The studio has ten ad formats and three models (the UGC ads page shows examples of each; for product-only openers without a presenter, the one-photo templates are an alternative):
| Model in the studio | Lengths | Credits, text start | Credits, photo start | Real-world completion rate* |
|---|---|---|---|---|
| Seedance 2.0 (720p, recommended, lip sync) | 5 / 10 / 15s | 12 / 24 / 36 | 12 / 24 / 36 | 79% (95% CI 60–90%), n=28 |
| VEO 3.1 Fast (720p) | 4 / 6 / 8s | 6 / 8 / 12 | 8s only: 12 | 98% (93–99%), n=123 |
| Kling 3.0 Pro (1080p) | 5 / 10 / 15s | 10 / 18 / 28 | 10 / 18 / 28 | 100% (87–100%), n=26 |
*Share of submitted jobs on all of Viralance (not only this studio) that produced a video, January 29 – October 1, 2026, from our real-world generation data. Observational: users chose their own prompts and settings, and samples are small (the Seedance 2.0 and Kling 3.0 Pro figures come from a handful of people, so treat them as indicative). Most Seedance 2.0 failures were policy refusals of what users asked for, not a measure of model strictness. Median waits were about 2–5 minutes.
Two details matter for hooks:
- In formats with a person on screen, the Hook field becomes the opening line they say to camera, and it is also saved as your caption. The Faceless format keeps it as the caption only. Generated speech can still change a word, so check the take. For on-screen text, use your ad tool, or burn in captions of the spoken line with Auto Subtitles on Edit Video (3 credits per render; it transcribes speech, so it needs a spoken hook).
- AI hooks costs 1 credit and returns six lines of 8–14 words, one per formula: Identity, Contrarian, Mistake, Curiosity, List, Story (refunded if the request fails). Rewrite anything that isn't true for your product.
Six hook formats for AI-generated UGC ads
Examples use a fictional mineral sunscreen; keep only claims you can substantiate.
1. Problem callout
What it is: name a specific, visible problem, then show the product. Why it can stop the scroll: it is self-relevant and states the proposition immediately, as TikTok asks. Studio format: Problem → Solution. Model: Seedance 2.0 when a person speaks; VEO 3.1 Fast for a cheaper draft.
- First line: "If your sunscreen leaves a white cast, look at this."
- Prompt: A young woman in a bright bathroom holds a small tube of mineral sunscreen to the camera and says: "If your sunscreen leaves a white cast, look at this." She rubs a small amount into the back of her hand. Handheld phone footage, natural daylight, vertical 9:16.
2. Result-first
What it is: open on the outcome. Why: it creates suspense and works without sound. Studio format: Before / After. Model: Kling 3.0 Pro or Seedance 2.0.
- First line (on screen): "Left: a typical mineral sunscreen. Right: ours."
- Prompt: Close-up of two hands side by side on a white towel, one with a visible white film and one with no visible residue, a small tube of mineral sunscreen between them. Slow push-in, natural window light, vertical 9:16.
Highest legal risk: a generated "after" is not evidence, and comparisons need substantiation (see compliance below).
3. Mid-action demo
What it is: frame one is the product in use; no greeting or logo card. Why: motion and product appear at once, which suits sound-off viewing. Studio format: Product Demo or Faceless. Model: VEO 3.1 Fast (cheap, high completion in our data) or Kling 3.0 Pro for 1080p detail.
- First line (on screen): "One pump, full face, 10 seconds."
- Prompt: Faceless close-up: hands press one pump of sunscreen from a white bottle onto fingertips and spread it across a cheek in one continuous motion. Product label in sharp focus, bright bathroom, handheld, vertical 9:16.
4. Identity callout
What it is: address one kind of person directly. Why: it filters for the audience you pay to reach. Studio format: GRWM or Tutorial, not Testimonial (see compliance). Model: Seedance 2.0.
- First line: "If you run outside at 6 a.m., this is the part of your routine to fix."
- Prompt: A young man in running clothes in a hallway looks at the camera and says: "If you run outside at six a.m., this is the part of your routine to fix." He picks up a sunscreen stick from a shelf. Morning light, handheld phone style, vertical 9:16.
5. Unboxing reveal
What it is: the reveal is the hook. Why: a small curiosity gap, with packaging on screen from frame one. Studio format: Unboxing. Model: Kling 3.0 Pro from a product photo (10 credits for 5 seconds); check packaging text against the photo (keeping labels readable).
- First line (on screen): "What $18 of mineral sunscreen actually looks like."
- Prompt: Hands open a small kraft box on a wooden desk and lift out the sunscreen tube shown in the reference image, turning it slowly toward the camera. Natural daylight, handheld, vertical 9:16.
6. Contrarian or mistake
What it is: challenge a common assumption. Why: surprise is one of the emotions TikTok names. Studio format: Tutorial or Comparison. Model: VEO 3.1 Fast or Seedance 2.0. Use it only for claims you can support.
- First line: "Mineral sunscreen doesn't have to make you look like a ghost."
- Prompt: A woman at a kitchen counter squeezes sunscreen along two fingers, rubs it into her cheek, then looks at the camera and says: "Mineral sunscreen doesn't have to make you look like a ghost." Bright kitchen, natural light, authentic UGC tutorial style, vertical 9:16.
A three-hook A/B test with real credit math
Do not use the studio's Variations (2 or 3) button for this. Variations 2 and 3 deliberately add "a slightly different camera angle and framing" and "a different background and lighting" to the prompt. That changes the picture as well as the hook. Instead, generate three single clips with the same format, model, length, persona, setting and vibe, editing only the quoted first line.
| Step | VEO 3.1 Fast, text start, 6s | Seedance 2.0, photo start, 5s |
|---|---|---|
| AI hooks (optional) | 1 | 1 |
| 3 clips, one per hook | 3 × 8 = 24 | 3 × 12 = 36 |
| Burn in captions for sound-off (3 × Auto Subtitles, spoken hooks only) | 9 | 9 |
| Total | 34 credits | 46 credits |
| Cost on Plus pack ($19 / 70 credits) | ≈ $9.23 | ≈ $12.49 |
| Cost on Business pack ($299 / 1,650) | ≈ $6.16 | ≈ $8.34 |
Without captions, the VEO plan is 25 credits and fits in the $9 Mini pack (30 credits). Keep credits spare for a re-roll: a hook with a garbled word is not a fair test, and completed clips you reject are still charged (refund rules for failed generations).
Running the test on the platforms
- Keep everything else identical: any footage after the hook, ad copy, landing page, audience, placement and budget.
- On TikTok, use a Split Test with creative as the variable. TikTok recommends running it for at least 7 days, choosing a budget that gives an estimated power value of at least 80%, and using a broad audience (TikTok split test best practices).
- On Meta, use an A/B test with one variable. Meta recommends running it for at least 2 weeks and not using the test audience in other campaigns at the same time (Meta A/B testing).
Reading the results
- Did the hook hold attention? Divide Meta's 3-second video plays or TikTok's 2-second and 6-second views (TikTok reporting metrics) by impressions. This isolates the opening.
- Did it sell? A hook can win attention and lose on cost per purchase. Choose the deciding metric before launch and wait for the platform's result, not an early lead.
- We give no benchmark "good" hook rate; we found no platform-published figure to cite.
Compliance: AI testimonials and AI labels
FTC (United States). Under the Endorsement Guides, ads that present people as "actual consumers" should use actual consumers or "clearly and conspicuously disclose that the persons in such advertisements are not actual consumers" (16 CFR 255.2). The 2023 revision clarified that endorsements include virtual influencers (FTC, June 29, 2023). The FTC's 2024 reviews rule also covers testimonials that misrepresent that they are "by someone who does not exist, such as AI-generated fake reviews, or who did not have actual experience" with the product, and allows civil penalties (FTC). In practice: don't script a generated person saying "I've used this for a month," don't show unsubstantiated generated results, and keep Testimonial lines about the product, not a claimed experience. This is not legal advice.
TikTok. AI-generated or significantly AI-modified images, video or audio in ads require a mandatory disclaimer; Ads Manager offers an AI-generated content disclaimer that adds a text label (TikTok ad disclaimers).
Meta. Meta labels ads made with its own generative AI tools (next to "Sponsored" when an AI-generated photorealistic person appears) and, per its June 1, 2026 update, will detect third-party AI ads through industry-standard signals and add an "AI info" label (Meta). Assume a UGC ad with a generated presenter may be labelled.
When not to use Viralance for this
- You need a real customer's testimonial. A generated person cannot give one.
- The ad depends on a real product result (skin, weight, cleaning). Film and document it.
- You want the tool to run the test. Viralance doesn't publish ads or report ad metrics; you upload the files to TikTok or Meta yourself.
- You need exact wording every time. Generated speech can change words; a Voiceover narration over faceless footage is easier to control.
Frequently asked questions
How long should a UGC ad hook be?
TikTok asks for the proposition within the first 3 seconds and the hook within the first 6. Write the first line so it can be said, or read as on-screen text, in about three seconds.
Can I label an AI presenter as a customer?
No. Under the FTC Endorsement Guides, people presented as actual consumers must be actual consumers, or the ad must clearly disclose that they are not.
Prices from Viralance's credit catalog and the UGC Ad Studio, checked September 26, 2026; see pricing. Completion data: Viralance production records, January 29 – October 1, 2026 (method and limits).
