That Influencer You Just Followed Might Not Even Be Real

Recently, I was scrolling through Instagram when a video showed up in my feed.

A bride, eye makeup smeared, sobbing into the camera while raving about an event-photo app called Once. She swore the app was how she captured every moment of her wedding — she was so worked up her eye makeup was a wreck.

A year ago, I would've shrugged: just another happy user doing an organic endorsement.

But that day, I held my gaze for three seconds.

The bride's face was too perfect. The emotion in her eyes ran too smoothly. The tears fell at exactly the angle that put the product logo squarely in frame.

I felt a chill creep up my spine.

I looked into it later. Reality Defender, a cybersecurity firm that specializes in deepfake detection, ruled that Once's ad was "highly likely" an AI-generated virtual character. The "bride" didn't exist at all.

And that's exactly the problem.

The person crying at you from the screen, pitching a product? She's never received a single shipment, never walked down an aisle, never shed a real tear in her life. She's just a piece of ad creative, engineered from scratch.

And you — you were treated as a real audience.

1

Some people might think this is a one-off.

It isn't.

Hyundai, Fenty Beauty, Prada, and Samsung have all, to varying degrees, used virtual characters in their campaigns. Lil Miquela, that CGI character you've probably scrolled past — her creator Isaac Bratzel went on to found a company called AvatarOS, dedicated entirely to building digital humans for brands.

This isn't dabbling. By some estimates, the virtual-influencer market will balloon to $45 billion by 2030.

Forty-five billion.

Why? Let me walk you through the math.

2

A traditional brand shoot — booking models, building sets, flying crews, hotel rooms, post-production retouching — runs at least $20,000, and as much as $70,000. Do that ten times a year and you're burning hundreds of thousands.

And a virtual influencer?

She doesn't call in sick. She's never late. She never complains the script is weak. She won't have a 2 a.m. meltdown on Twitter. She won't fight her agent over pay. And she won't suddenly announce she's off to find herself on a spiritual retreat.

She can roll cameras 24/7, get translated into dozens of languages, and be re-outfitted, re-faced, or re-scene with a single click. Today you have her pitch coffee; tomorrow she's the face of a car.

This isn't an influencer. This is a content machine you can recycle infinitely.

Capital can smell the money. In March 2025, AvatarOS closed a $7 million seed round led by M13, with a16z's Games Fund following on. In January 2026, Synthesia, a London-based AI digital-human company, raised $200 million at a $4 billion valuation.

Strictly speaking, Synthesia isn't an influencer company — they make enterprise training videos. But look at that valuation, and you'll understand what investors are betting on.

The bet is simple: real people are too expensive, fake ones are cheap.

3

But here's the thing: this math can't be done in money alone.

That Once bride ad? Nowhere does it say "AI-generated." The video was pushed into real users' feeds, indistinguishable from a genuine customer's selfie review.

Tell me — how are you supposed to tell?

You can't, because these tools were designed so you can't.

There's a survey that hits hard. Which?, the UK consumer organization, ran a deepfake test: they mixed real videos with fake ones and showed them to ordinary people. The result: seven in ten couldn't get every answer right.

Three in ten didn't even realize they'd been duped.

So the core contradiction comes into focus:

Every cent a brand saves is withdrawn from your trust. You place an order because you believe "this is a real user's experience." The moment that premise turns out to be fake, the entire foundation of the transaction collapses.

4

And regulation?

I went and looked. Four words sum it up: still on the road.

In the US, the FTC (Federal Trade Commission) updated its endorsement guidelines in 2023, requiring paid promotions to be clearly labeled and banning deceptive endorsements. But those guidelines were designed for "real influencers forgetting to tag #ad." They didn't anticipate a new scenario: the person doing the endorsing doesn't exist at all.

The EU went harder. The AI Act includes a provision that, starting in August 2026, requires all AI-generated or AI-altered content to carry a machine-readable label, so platforms and regulators can identify it.

But enforcement only just began this month. Every AI ad that ran before that is still floating around in feeds, untouched.

As for the UK, the ASA (Advertising Standards Authority) has admitted that current rules don't explicitly ban the practice of running ads without an AI label — they can only say "ads must not mislead consumers." Which amounts to saying nothing at all.

The rules are chasing, but they'll never catch the speed of the feed.

5

Let me be blunt: the brands going hardest on AI influencers right now are essentially making a bet.

They're betting on two things:

First, before the regulatory hammer falls, lock in ROI at low cost.

Second, once AI ads become the norm, consumers will go numb, forgive it, and decide "everyone's doing it, no big deal."

The first bet, they might win.

The second one, I seriously doubt.

Why?

Because consumers can tolerate small slips like "an influencer didn't tag an ad." What they can't stomach is "I was talked into buying by someone who doesn't even exist." The first is a process problem. The second is a matter of dignity.

Picture it: someone opens the video full of trust, watches the "bride" tearfully recommend a wedding product, places the order — and then discovers the bride was generated by code. How does that person feel?

They feel like they were played for a fool.

And there is nothing people resent more than being played for a fool.

So my read is this: the first major brand to be publicly exposed for undisclosed AI virtual spokespeople will be pilloried for a long time. Nobody will remember how much they saved or how pretty their ROI looked. They'll remember exactly one thing: it made me believe in a person who never existed.

The cost of repairing that trust is vastly more expensive than the few ten-thousand-dollar shoot fees they saved.

6

I'm not against AI influencers.

Honestly, if a brand openly labels something as "our virtual spokesperson," that's just another content format — like Hatsune Miku, or how Lil Miquela started out. The audience knows it's fake, and they pay because they're buying into the creativity and the aesthetics.

What I oppose is disguising a fake as a real person, then charging real people money.

This isn't a technical problem. It's a question of attitude.

That EU transparency rule taking effect in August is, in a sense, hitting the brakes for the entire industry. If the brands won't hit the brakes themselves, somebody will hit them for you sooner or later.

And when that happens, the bill will be far steeper than it is today.

I don't know how this all ends. But there's one question every brand still wavering on whether to deploy an AI influencer should sit down and think about:

Are you willing to be the one your trusting audience discovers was fake?

Flip the perspective, and it gets clearer.

Next time a "real user" crying into the camera to recommend a product shows up in your feed, pause for three seconds — and look at her eyes.