Our AI influencers' first product ads were 12 to 14 seconds long, almost silent, and beautifully lit. Every one of them felt fake. The owner's verdict was blunt, and fair: no people around, no background noise, no naturalness, an idea too thin to care about.
So instead of tweaking prompts, we went and studied what works. We collected 23 creator videos about small gadgets and wearables on TikTok, each with over 100,000 views, and read their length, sound and spoken words from the platform's own page data. Then we downloaded the strongest ones and went through them frame by frame, with full transcripts and sound levels. This is what they have in common.
Frame by frame, the winners look nothing like a polished ad.
What we looked at
23 creator videos about translator earbuds, smart glasses, a smart ring, AI voice recorders, a mini projector, a toothbrush and other gadgets, all above 100,000 views on TikTok, checked on October 27, 2026.
The strongest ones studied frame by frame: a sponsored electric toothbrush video (2.9M views), a two-person translator earbud skit (35.5M), a paid smart-glasses video shot at a farmers market (13.1M), a smart-ring story told to camera (2.5M) and a sponsored voice-recorder video (about 989,000).
For contrast, a set of weaker videos about the same products: product close-ups over music, with nobody talking.
It's a small sample and view counts move, so treat the patterns as patterns, not laws. But they were consistent enough to change how we make ads.
Most disappointing UGC videos start with a weak brief: a product link and 'make something fun'. Here's a one-page brief template for product videos, field by field, with what to include, what to leave to the creator, and the rules to put in writing.
20 of the 23 videos have a person speaking, and 21 of 23 use their own audio rather than a music track laid on top. The toothbrush video is 43 seconds of continuous first-person talk, roughly 110 words. The earbud skit is a conversation from start to finish. Even the smart-glasses video, which is mostly first-person footage, has the creator's voice over all of it.
The words do the selling: what happened, what the product did, what it costs, where to get it. (We turned this into a five-beat UGC ad script template.) Pictures alone, however good, didn't carry an ad in this set. The product-montage-over-music videos about the very same gadgets sat between a few hundred and about 13,000 views.
2. They open on a situation, not the product
The first line is almost never the product's name. It's a moment:
the smart-ring video opens with the creator's belief that the ring may have saved their life the day before;
the earbud skit opens in a parked car after a nail-salon visit, with "what did they say in there?";
the toothbrush video opens with a strong personal verdict: the most futuristic toothbrush the creator has ever used.
The product arrives as the explanation of the story. A stranger stays because they want to know what happened, not because they want to hear about a feature.
3. Ordinary places, with other people in them
Nail salons, parking lots, a farmers market, a bathroom, a car. Not one studio. The smart-glasses video is around 40 short shots of a busy market: vendors, a handover, a dog, other shoppers. The highest-viewed video in the whole set, at 35.5 million views, is two friends in a car asking each other "where'd you get those?", "how much?".
Other people do two jobs: they make the place feel real, and they ask the viewer's questions out loud. That second-person "what is that?" is one of the most effective lines in product content, and it costs nothing.
4. They're longer than you think, and they cut constantly
Most of the hits run 30 to 60 seconds, and the typical length in our set was around 50 (for paid ads, see how long a TikTok ad should be). None of them feels slow, because something changes every few seconds: a new angle, a close-up of the product doing its job, a cut back to the face. The toothbrush video has about 30 shots in 43 seconds, and the creator's face is on screen for only about a third of them. The rest are hands and the product: the box, the light ring, the pressure warning, the app.
Short isn't the goal. Never boring is.
5. Real sound, all the way through
The sound in these videos never drops out: the voice on top, and under it the room, the street, the car, the product's click. In our own silent ads, the quiet between lines was one of the clearest signs that something was off. Real places are never silent.
A phone on the dashboard: one of the most common, and most believable, setups.
What the weak ones do
The videos that didn't travel were, almost word for word, what our first AI ads looked like:
the product on a clean surface, slow camera moves, music, no person;
a list of features read over b-roll, with nothing happening to anyone;
polish everywhere, life nowhere.
One more finding from the data: the same creator posted near-identical versions of one earbud video, and one cut reached 25 million views while others stayed around 30,000. The hook and the first seconds matter more than anything else in the video, so it's worth making more than one version.
What we changed
We rebuilt how our AI influencers make ads around these patterns:
A spoken story first. The ad is written as something she says out loud, 25 to 40 seconds: a situation hook, what happened, the product as the answer, one or two features shown working, an honest close ("yes, this is an ad") and the call to action.
Her voice first, pictures second. Her lines are recorded first, and the video is timed to them, the way an editor cuts to a voiceover.
Talking shots plus real b-roll. Shots of her talking to a phone on the dashboard or a kitchen counter, lip-synced, cut with b-roll of real places with people in the background.
Sound under everything. The car, the street, the laundromat: every shot has the sound of its place.
Every clip checked. We check that the face stays the same person throughout and that no clip "says" words nobody wrote, both things AI video gets wrong surprisingly often.
The first ad made this way, about a small wearable that answers questions out loud, was a different thing entirely: a 40-second story set in a laundromat, with other people in the background, about shrinking a jumper. It wasn't perfect (the pacing still needed more cuts), but it looked like something a person would post.
If you make product videos, AI or not
Write the first line before anything else, and make it a situation, not a product name.
Talk. In your own voice, specifically, like you'd tell a friend.
Film where the problem actually happens, and let other people be in it.
Show the product doing its one job, in close-up, more than once.
Cut every few seconds, and keep the real sound.
Say the price or where to get it, and own that it's an ad.
Influgen's AI influencers now make ads this way: her ad manager researches the product, her producer writes the whole spoken story and shot plan, and the platform records her voice, films each shot from approved stills, checks every clip and edits it together. You approve the stills before anything expensive is made, and the video before it's posted. It works for tech, beauty, fitness and every other niche. See the pricing, or create your influencer for free now; their team starts when your spot opens.
A good AI UGC ad is decided before anything is generated: in about 60 words of script. Here is the five-beat structure our AI influencers' producer writes to, how many words fit in 20 to 30 seconds, and the mistakes that make a script sound like an ad.
Most weak product ads aren't badly filmed. They're made by someone who never learned how the product works. Here is the research good creators and agencies do before writing a single line, and what we learned building it into our own pipeline.
