A casting table covered in headshot cards of ordinary people, with four pulled forward and ticked
Scroll past enough product ads and a type starts to repeat. The woman in her early thirties holding a serum bottle up to a bathroom window, talking a little too fast. The guy in a hoodie at a desk with two monitors and an RGB keyboard, explaining why his old headset was ruining his aim. The mom in a kitchen with a toddler just out of frame, who found a lunchbox that finally fits in the backpack.
None of them look like actors. That is the entire point. And almost none of them got there by accident.
The format is called user generated content, and the name has not been accurate for a while. What began as customers filming themselves has become a production category with rate cards, agencies, briefs, and above all casting. The person who looks like they wandered into frame was usually chosen from a shortlist, for reasons that would be familiar to anyone who has ever sat in on a casting session for a film.
Character actors, not leads
Film casting has always split into two jobs. Leads are cast for presence, for the quality that makes an audience willing to follow someone for two hours. Character actors are cast for recognition, for the ability to walk on screen and make the audience believe instantly that this is the landlord, the night nurse, the bartender who has seen everything.
Nobody watching a character actor thinks about how attractive they are. They think the bartender looks like a bartender, and the scene moves on. The best character actors are almost invisible as performers because they are so completely legible as types.
UGC advertising runs almost entirely on the second job. A skincare brand does not want a model, because models read as advertising and advertising is the thing the viewer has learned to skip. It wants somebody whose skin looks like skin, with a bit of texture, somebody who plausibly has the problem the product fixes. A protein powder wants someone who visibly trains but does not look like a fitness influencer. A gaming headset wants a twenty-two year old with a desk setup that is slightly too elaborate and a room that is slightly too dark.
Get the type right and the viewer’s brain files the clip as a person talking rather than a brand talking, which buys three or four seconds of attention. Get it wrong, by casting someone too polished or simply unsuited to the product, and the thumb keeps moving before the hook has finished.
The audition tape nobody sees
Brands learned this the hard way. The early rush into creator content produced a lot of expensive clips in which the creator was talented, likable, and completely wrong for the product. A twenty-year-old recommending retirement planning software. A glamorous lifestyle creator selling budget cookware. The content performed like advertising because the casting said advertising before anyone spoke.
So the process tightened. Agencies started holding rosters organised by type rather than by follower count. Briefs started specifying age range, setting, energy, and sometimes wardrobe, the way a casting call would. Creators began tagging their own portfolios with the roles they could credibly play.
What emerged looks a lot like a small casting industry, except the auditions are mostly performance data. A creator who reads as a believable new parent gets booked again when the parenting products convert. One who reads as a believable gamer gets the next peripheral launch. The market is quietly sorting people into the parts they can play.
That also made it slow and expensive in exactly the way casting always is. Finding the right face for a single product means sourcing, briefing, waiting for delivery, and hoping. Testing whether a different type would have worked better means doing all of it again.
A phone on a tripod showing eight different people holding the same product, like screen tests for one role
Casting as a search box
This is where the category changed shape. Tools built for ai ugc ads take the casting step and turn it into something closer to browsing a directory. UGCfy AI, for instance, keeps a library of around three hundred synthetic presenters that a marketer filters by fit for the product and the audience, then pairs with a script generated from the product page itself.
The synthetic performer gets the headlines. The bigger shift is what happens to testing once casting stops being the bottleneck. A brand that used to commission one creator and hope can now run the same script through a skeptical thirty-five year old, an enthusiastic twenty-four year old and a tired parent, see which type the audience actually believes, and learn something about its customers that no brief would have told it.
For anyone who has followed how films get cast, that is a genuinely strange development. Screen tests used to be the most expensive part of finding the right face. Here they cost almost nothing, which means the question of who is believable for which product gets answered by the audience rather than by somebody’s instinct in a meeting.
What the strike was actually arguing about
Anyone who followed the 2023 actors’ strike will hear an alarm going off at this point, and it should be examined rather than waved away.
The core fight over digital replicas was about consent and compensation for real performers. Studios wanted the ability to scan a background actor once and reuse that likeness indefinitely. The union won requirements for informed consent and payment when a real person’s likeness is recreated. That principle is the right one, and it is the one worth holding any synthetic performer to.
The honest test for an AI presenter is where the face came from. A presenter built from a performer who agreed to be modelled and was paid for it sits on one side of that line. A face generated to resemble nobody in particular sits somewhere else again. A face that recreates an identifiable person without their agreement is exactly what the strike was about, and no advertising efficiency justifies it. Buyers are entitled to ask which of those a tool is offering, and the tools that answer clearly are the ones worth using.
There is a disclosure question too. Several platforms and a growing number of regulators expect audiences to be told when the person speaking in an ad never used the product, and that expectation is only going to get firmer. A brand that treats disclosure as an obstacle is misreading where this is heading.
The skill that does not get automated
What survives all of this is the part that was always hard: knowing which type fits which product. A tool can offer three hundred faces. It cannot tell you that your ergonomic chair sells to people who have already had a back problem, and that the right presenter is forty, slightly rumpled and faintly annoyed rather than young, energetic and fine.
That judgment is casting, and casting directors have been doing it for a century. The UGC market has simply rediscovered it, a few years late, and given it a much faster screen test. The random person in the next ad you scroll past is still a choice somebody made. The difference now is how many other people they tried first.
