Pretty much every week, a brand marketer asks me some version of the same question: should we be doing AI UGC?
And every time, the demo they've been shown is the same thing: an AI avatar (sometimes a licensed real person's face) reading a script that's framed as a customer testimonial, even though no customer ever said a word of it.
At some point the whole category just decided that "AI UGC" means fake people. I build Bevyl, so I'm in this market every day, and I think that's the wrong bet. Here's why.
"AI UGC" means two totally different things
Ask ten tools what AI UGC is and nine of them will show you an avatar. But there are actually two versions of the idea:
- Generated UGC. An AI actor performs your script. You get a "testimonial" in minutes without ever shipping product to anyone.
- AI-edited UGC. Real customers and real founders shot the footage. The AI just does the editing, which is the part that was actually slowing you down.
The first version gets all the demos. I think the second one is the better long-term bet.
Why the avatar version works right now
I'm not going to sit here and tell you avatar UGC doesn't convert, because right now it often does. It's cheap and fast, and you can test a bunch of scripts before lunch.
But think about why UGC converts in the first place. A shaky phone clip beats a polished studio ad because the viewer reads it as a real person's experience instead of an ad. An avatar testimonial looks like UGC, but there's no actual customer behind it, and the assumption that there is one is the entire reason the format converts.
That only works while viewers can't spot it. And they're getting better at spotting it:
- TikTok requires creators to label realistic AI-generated content, and Meta labels AI-generated media across its apps. Those labels are only going to get bigger and harder to miss.
- The FTC's endorsement guides are pretty clear that testimonials need to reflect real experiences from real users.
- People call this stuff out in the comments constantly. One "this is AI" comment with a few hundred likes and the ad starts working against you.
- Also, something the demos skip: other brands are renting the exact same faces. The same "customer" raving about your product is running in their ads too, and your audience scrolls past all of them.
If you're running avatar ads today, fine, but know that this stuff has a shelf life. Watch your comments and watch how fast each creative burns out.
My prediction (hold me to it)
Brands still running undisclosed avatar testimonials in mid-2027 will be paying for it in CAC. Their creative will burn out faster and the comments will call it out more often.
If avatar UGC still converts like it does today by then, I was wrong and you can quote this post at me.
So what should you do instead?
What people forget about real UGC: the performance was never the issue. It was a pain to produce. You had to source the clips, chase the permissions, and then someone had to edit all of it, and that last part is where most teams gave up.
Editing is the part AI actually solved. You don't need AI people, you need something that gets through the pile of footage your customers send you.
Quick gut check before the pipeline: out of your last ten UGC-style videos, how many have an actual customer in them? Hold that number.
The pipeline looks like this:
- Get real clips for free. Post-purchase requests, founder phone footage, reposting customer tags. We wrote the whole playbook in UGC-style videos without hiring creators, including the 3-prompt script that gets customers to send footage you can actually use.
- Don't over-edit what comes in. The shaky camera, the bedroom lighting, the person fumbling a word, that's the stuff viewers trust. Trim the clip, add captions, and leave the rest alone.
- Let AI handle the volume. This is where teams were losing 3-5+ hours per video, and it's the one step you can automate without touching what makes the footage believable.
We built Bevyl to do exactly that third step and nothing else. Train it on your brand (takes about 5 minutes), then feed it the customer clips and founder footage you already have, and it edits them into on-brand shorts. It doesn't add avatars or invented footage: the person your viewers see is always your real customer. People ask us when we're adding AI actors. We're not, on purpose.
What to do this week
Two moves, in order.
First, run that gut check from earlier and get the real number. You might be surprised how few actual customers are in there, and better to find out now.
Second, start collecting real footage. Turn on a post-purchase clip request, pull the customer tags you already have into one folder, and let it fill up. When editing becomes the bottleneck, AI belongs in your UGC stack. Upload the folder to Bevyl and see what a week of real customer footage turns into, or talk to the team first if you want a second opinion on what you've collected.