Native beats polished, and it is not close
The instinct to make it look expensive is the thing that kills performance. A handheld clip with imperfect lighting matches what the viewer was already watching, so the scroll does not break; a colour-graded brand film announces itself as an ad in the first frame. Match the feed, not the brand guidelines.
- Shoot vertical, handheld, in a real room. A studio background reads as paid
- Keep the first frame unstaged: a face, a screen, a mess, not a logo
- Let the audio be a voice, not a music bed
- Captions always, because most of the feed is watched muted
The script skeleton
Almost every UGC video that works runs the same five beats, in the same order, inside 30 seconds. Write to the skeleton and the only creative decision left is the hook, which is where the effort belongs.
- 1Hook, in under three seconds: the problem or the objection, stated as the viewer would state it
- 2Agitate briefly: why the obvious fix does not work, in one sentence
- 3Show the product doing the thing, on screen, not described
- 4Give the result in specifics: what changed, how long it took
- 5One CTA, spoken and captioned, naming the next step exactly
Cut every sentence that does not move one of those five beats. A 20 second video that lands all five beats outperforms a 60 second one that lands three and explains your founding story.
The hook is the whole job
Everything after the first three seconds is only seen by the people the hook kept. Write ten hooks per script and shoot the same body against all of them, because the hook is the cheapest thing to change and the only one that reliably moves reach.
- Lead with the objection: "I thought AI video looked fake too, until this"
- Lead with the specific: a number, a timeframe, a named tool
- Lead with the mistake: "You are posting daily and getting nothing. Here is why"
- Lead with the demo: start mid-action, on the screen, with no introduction
- Never open with your name, your company, or the word welcome
Human creators, AI creators, or both
Human creators bring a face people already trust and an audience of their own; they also bring sourcing, briefing, revenue share and a two-week turnaround per variant. AI creators bring volume: the same script in ten hooks, in a day, for the cost of the tool. Most teams end up using both, for different jobs.
- Use human creators when the trust of a specific face is the point, or when the creator's own audience is part of the deal
- Use AI creators for volume testing, for localisation, and for the weeks nobody can film
- Keep one consistent AI character rather than a new stranger each time, so the account builds familiarity
- Test both against each other in ads before deciding, because the answer varies by category
The Influencer marketing chapter covers finding, briefing and paying human creators. The mistake to avoid here is treating AI UGC as a cheaper clone of that: its advantage is iteration speed, so use it to test twenty hooks, not to make one video slightly cheaper.
Disclose it, both kinds
Two separate rules apply and they catch people out for different reasons. Paid creator content must be disclosed as an ad, and AI-generated or AI-altered content must be labelled on most platforms. Neither is optional and neither meaningfully hurts performance; getting caught does.
- Anything you paid for, gifted or affiliated needs a clear disclosure the viewer sees, not one buried in hashtags
- Use the platform's own paid-partnership and AI-content toggles, not just a line in the caption
- Never use a real person's face or voice without written permission, including a customer's
- Do not have an AI creator claim personal experience of a result they did not have
The disclosure rules differ by country and platform and they change. Check the current guidance where your audience actually is, and when in doubt over-disclose: an obvious label costs a little reach, an undisclosed ad costs the account.
Volume is the testing strategy
No one picks the winning hook by taste, including you. Ship many variants of the same script, let a week of data pick, then make ten more of whatever won. This is the same loop as the Paid ads chapter, run organically first because the feedback is free.
- One script, five to ten hooks, posted across the week
- Judge on three-second retention and saves, not likes
- Rebuild the winners with a new hook rather than a new idea
- Retire a hook when its retention drops two runs in a row, not the first time it dips
Key moves
0/7Reading is warm-up. Check these off as you actually ship them; progress saves in your browser.
Ten hooks by tomorrow, not ten days
Viraloop generates UGC-style videos from your website with a consistent AI creator, so testing twenty hooks costs an afternoon instead of a shoot.