If you've ever run UGC ads, you know the treadmill. A creative works, you scale it, and three weeks later frequency is up, click-through is down, and you need a fresh batch. The brands that win are honestly just the ones that keep the fresh batches coming.
That treadmill is why "AI UGC ads" became a category. And almost every tool in it sells the same shortcut: AI-generated people instead of real customers.
There's a second way to do this, where you use real footage and the AI just handles the editing. It's the version I'd bet a paid account on. Here's the playbook.
So which route should you take?
Route one: AI avatars. Arcads, Creatify, and HeyGen all sell UGC-style ads fronted by AI avatars. The pitch, at least on their sites, is speed: script to video in minutes, dozens of variants, no creators to wrangle.
What the pricing page doesn't mention: you're licensing those faces, not hiring the people, and the license isn't exclusive, so the actor you picked can front a supplement brand and a mattress brand in the same week. The AI-content labels keep getting bigger too. And paying to promote an ad where the top comment says "this is AI" is a terrible use of ad spend. I wrote a whole post on this: AI UGC in 2026: real footage beats fake faces.
Route two: real footage, AI editing. Your customers, your founder, and your creators shoot the clips. Customer and founder clips cost you nothing but the ask, creator clips carry a fee, and either way the person on camera is real and holding your actual product.
Route one gets you an ad faster this week. But you'll be running this program for quarters, and over that stretch route two ends up cheaper and more believable.
The playbook
1. Build the free footage pipeline
Post-purchase clip requests, founder phone footage, reposted customer tags. The full sourcing system is in UGC-style videos without hiring creators. One extra step for paid: get usage rights for ads in writing when you collect the clip, not after the ad is live.
2. Brief with three prompts, not a script
Hand a customer a script and you get an actor, which defeats the whole purpose. Use the 3-prompt brief instead: show it in your hand, say one thing that surprised you, show the result. Picking the opening before anyone films is the same argument the 3-hook framework makes for the edit.
3. Let AI do the editing
Okay, so footage is coming in. Now the part everyone thinks will be quick.
A real TikTok or Reels test wants fresh variants every week, and manual editing is usually where that cadence falls apart. Teams tell us they were spending 3-5+ hours per video before Bevyl.
This is the AI part of the AI UGC ad, and honestly it's the only part that should be AI. Bevyl trains on your brand in about 5 minutes, then edits raw customer clips into on-brand shorts at whatever pace your testing needs. It doesn't generate people, which matters for compliance, not just taste. Every face in your ad is a real person your brand actually works with: a customer, your founder, or a creator you briefed.
4. Get five ads out of one clip
Any media buyer will tell you the first two seconds are most of the ad. So take each strong clip and cut several different opens onto the same body. Our TikTok hooks breakdown has the patterns. You'll get more from five hooks on one strong clip than from five new mediocre clips. And it's way less work.
5. Ship through Spark Ads
Whitelisted posts keep the native look that made the footage work in the first place. Our TikTok Spark Ads guide covers the setup, and TikTok ads cost covers what to budget.
The cost comparison
Run your own numbers, but here's the rough version. On the avatar route you're paying per generated video, and a real testing program wants dozens of variants a month, so the bill grows every time you test more, which is the whole point of testing. On the real-footage route the big costs are one-time: say a creator clip runs you somewhere in the low hundreds with usage rights, and customer and founder clips are free. Every extra hook variant off the same clip is basically just another editing pass.
And launch-day click-through only tells you who started strong. What decides winners is how long the creative keeps working and what the comments look like under it. Footage of actual customers in their actual bathrooms holds up better on both.
Where to start
Turn on a post-purchase clip request today. Footage takes a few weeks to start coming in, so start collecting before you need it.
Then grab the five best clips you already have, run them through Bevyl for hook variants, and put them behind a small Spark Ads test next week. Want a second pair of eyes on your footage mix first? Talk to the team.