Thinking · Newsletter
UGC ads for performance: the complete guide (actors, creators, AI, and the organic brief)
UGC versus produced by job, why we cast actors not influencers, casting to a role, the four meanings of organic, where AI wins and humans still do, and why the edit is where volume comes from.
TL;DR. UGC is a production method, not a look. It is good at believability, speed and volume, which makes it the right default for testing messages at scale; produced work is right where cheapness reads as risk or where you need to stand out in a feed that is now all UGC. We cast paid actors and creators rather than influencers for performance work because a test needs the same person saying different things, and clean rights. We run generative video in the same pipeline and keep a roster of more than eighteen hundred rights-cleared human creators, because the two do different jobs: the models win on single-speaker monologue, camera moves and the fifteenth variant; humans win wherever the ad has to be believed as testimony or the product has to be handled. "Make it feel organic" is four different requests, and only one of them matters. Here is the whole practice in one place.
UGC or produced: pick by job, not by fashion
| UGC | Produced | |
|---|---|---|
| Best for | Testing message at volume | Differentiation, trust, launches |
| Cost per asset | Low | High |
| Assets per shoot | Many | Few |
| Fails when | The category needs authority | You need thirty by Friday |
Most accounts should run mostly UGC and a little produced; most accounts run all UGC. The mix is the decision. Running one format exclusively is how a library becomes invisible in a feed where everyone made the same choice. Category changes the answer: in apparel, how the thing looks is the argument and produced beats UGC more often than anywhere else; in supplements and finance, a phone camera can read as a red flag where the buyer is already sceptical. See creative conventions by vertical.
Actors and creators, not influencers
For performance work we book a roster of paid actors and creators rather than influencers, for three practical reasons. Control of the script: an influencer's voice is the point, so the script bends to them, and controlled comparison becomes impossible. Repeatability: a test needs the same person delivering different messages or different people delivering the same one, and a roster can be scheduled that way. Rights: paid creators sign; usage and duration are clean, and nothing is renegotiated when an asset works and you want six more months of it. Influencers are right when the audience is the asset: a launch, a credibility problem, a category where one person's endorsement changes the decision. That is real and it is not performance creative. Full reasoning in why we work with actors.
Cast to a role, not a demographic
"Woman, 30s" produces a performance nobody can direct. "Someone who has tried three things that did not work and is sceptical about a fourth" produces a scene. The same substitution fixes the brief, the casting call and the direction on the day, and it makes the result reusable: if scepticism-after-failed-alternatives beats wanting-a-simpler-routine, that finding carries into the next brief whoever it reached. See brief to motivation, not to demographics.
"Make it feel organic" is four requests
Clients ask for organic more than any other note and it is the note we have got wrong most often, because it means four things. Unpolished (phone camera, imperfect light) is a production instruction. Unscripted (sounds like thinking, not reciting) is a performance instruction. Unbranded (no logo first) is a structural instruction. Believable (this person seems to use it) is a different problem from the other three and the only one that matters. You can deliver the first three and miss the fourth: a grainy, unscripted, logo-free ad from someone who has clearly never touched the product is less believable than a well-lit scripted one from someone who has. We now ask which of the four, and what would make this person credible to your buyer.
Where AI wins and where actors still do
We are AI-native and we still pay actors. After a year of running both in production the line is fairly clear.
Generative wins: single-speaker monologue to camera (lipsync and range on one person are the strongest thing the current models do); camera movement (pans and tilts come out clean with little coaxing); variation at the edges (the fifteenth framing of a working concept is close to free); and turning an owned product still into b-roll for a test that has not earned a shoot day yet.
Humans win: anything that has to be believed as testimony; hands, products and physical interaction (held, opened, applied, worn); multi-person scenes; continuity across shots. Booksy, a marketplace app, reported in October 2026 that fully AI ads "significantly underperform" AI-edited ads featuring real providers, and that its audience has a measurable aversion to AI-made content. AppLovin's CEO said the same month that the models cannot yet make a good 30 to 60 second clip. Our line: no synthetic replicas of real people and no fabricated testimony. If an ad implies a person exists, that person exists, was paid and signed. The full account is in AI UGC: where it wins, where actors still do and what Veo3 is actually good at.
The edit is where the volume comes from
One shoot day with a cast performer yields many finished assets only if the edit is planned for it: multiple openings shot, two angles per beat so the edit has options, captions and titles and pacing varied in post. Booksy described extracting "30, 40 different versions... actually hundreds" from one piece of creator content. Volume lives in the edit, not in more shoot days. See the edit is where creative volume is won or lost and the pre-flight checklist for a batch.
Sound is half the ad
A viewer with sound off gets one ad; with sound on gets a different one, and on most placements a meaningful share are in the second group. Three mistakes: music chosen last, so it fights the cut; voice mixed too quietly against a bed; and no silence, when a beat of nothing is the loudest thing in a noisy feed. Cheapest test in creative: the same cut with voice-led, music-led and a deliberate pause in the first second. Captions are not optional on any placement where the sound is off by default, which includes most in-game inventory.
What UGC cannot survive
A destination that changes the subject. A believable creator explaining exactly why the product suits this buyer, landing on a product grid, is a wasted performance however it was shot. The ad and the page are one test. See landing page vs product page for paid ads.
Questions people ask
What is UGC advertising? Ads shot in the register of a user's own content, usually one person to camera on a phone, made by paid creators or actors for the advertiser. In performance work it is a production method chosen for believability, speed and volume.
Should I use influencers or UGC creators? Creators and actors for testing messages at volume with clean rights; influencers when you are buying their audience rather than their performance.
Does AI UGC work? For monologue, camera moves and variants, yes. For testimony, product handling and anything the audience must believe happened, cast a human. Booksy's and AppLovin's public statements in October 2026 both point the same way.
How many UGC ads do I need? Enough for the network to learn from. On AppLovin the published working number is 182 active creatives per top campaign; on Meta a smaller set carries, but the opening should still be varied five ways per concept.
What does "make it organic" mean? Unpolished, unscripted, unbranded or believable. Ask which; only the last one moves performance.
Sources: Flixr production notes, 2026; Gamesforum, "Inside Booksy's Creative Playbook", October 2026; Northbeam, "Inside the Rise of AppLovin", October 2026; AppLovin creative-set data, June 2026.