That "Bride" You Scrolled Past Might Not Exist at All

Recently I scrolled past a video.

A bride, in her wedding dress, crying at her own wedding. She said she'd originally planned a "phone-free wedding," but changed her mind at the last minute and handed every guest a disposable camera. The camera panned, and in the corner of the frame an app name popped up: Once.

The comments section was a flood of emotion. "This is what a wedding should be." "So thoughtful."

I almost believed it too.

That is, until I saw the second video, and the third… the same bride, the same spiel, only the camera angle had changed. That's when it hit me: this "bride" might not even be human.

What Exactly Is an AI Influencer?

Put plainly, it's when a brand pays someone to use AI to generate a face, a voice, a video, and have it pretend to be a real customer doing a "personal recommendation" on the brand's own social media. In other words, an AI-generated influencer — a fake person engineered to look and sound real.

It's not the kind of obviously cartoonish virtual avatar you see streaming. What it's after is indistinguishable from the real thing. An ordinary person's face, an ordinary person's tone, an ordinary person's everyday life — engineered so you'd swear it's just the neighbor next door sharing a product they love.

Good lord.

And it's not just me feeling this way. Two security firms that specialize in deepfake detection — Reality Defenders and Get Real Labs — ran an analysis and found that the promotion of the Once app was, in all likelihood, using exactly this kind of AI influencer. I reached out to Once. They didn't reply.

But This Goes Far Beyond Once

So I kept digging.

There's an app called Maket that uses AI for home design. On its Instagram, a girl's post reads, "I could kiss the insider who showed me these floor plans." Then she whips out her phone and demonstrates how Maket works.

Maket did reply, actually. Their answer was candid: AI influencers are just one of several tools they use for small-scale creative testing. "It's not the core of our marketing — more like an experiment, to figure out which hooks land best across different channels."

Then there's a Dubai fashion brand called Ashle. They posted a photo of a girl wearing their clothes at a restaurant. Everything looked perfect — except her hand. She had one too many fingers. The classic AI-image meltdown. When asked about it, Ashle took the photo down and explained that the clothes were genuine, hand-tailored pieces; the AI imagery had only been used in early promotion, and the posts were pulled because those styles had since been discontinued.

See? Excuses always arrive before the truth.

Regulation? Yes — But With a Giant Hole

At this point you're bound to ask: is nobody regulating this?

The EU has stepped in. Its AI Act began taking effect in August, requiring that AI-generated or AI-altered content — deepfake images, audio, video — be clearly labeled.

But this law doesn't apply in the UK.

The UK's Advertising Standards Authority (ASA) stated plainly: "None of our rules prohibit brands from publishing AI-generated promotional content, nor is there any requirement to label AI." Their only red line is a single one: advertisements must not mislead. As for AI itself, they don't scrutinize it. "What we look at is whether an ad leaves consumers with a false impression — not whether the AI was used correctly."

To put it bluntly: as long as you don't get caught, no one stops you.

Lisa Barber, tech editor at the consumer organization Which?, shared a set of numbers that sent a chill down my spine. They ran a test, showing ordinary people a mix of real and fake videos. 70% of people couldn't identify every fake video. In other words, seven out of every ten people are being deceived — and haven't a clue.

"Consumers can no longer believe what they see online," she said.

Who's Manufacturing These "People"?

I followed the thread further and turned up someone named Clarissa Mansbridge. She used to work in celebrity management — she represented Katie Price — and now runs a brand called Mia Metaverse, which custom-builds AI influencers for brands.

She did the math for me.

For a brand to shoot a set of high-end, campaign-grade images the traditional way, it runs $20,000 to $70,000. Photographer, models, location, post-production — line item after line item. Swap in an AI influencer? The cost gets slashed, and it simultaneously does away with the things brands dread most: no negative press, no personal opinions blowing up in your face, no hourly-billed temper tantrums, no photographer.

Mansbridge put it with brutal directness: "The influencers killed their own market. Brands wanted that 'real ordinary person using the product' kind of content, so influencers churned it out relentlessly. And then brands realized — if what we're after is authenticity, well, AI can be 'authentic' too."

She even tossed out a number, which I double-checked twice: by her estimate, 40% to 60% of the content put out by some major brands is AI-made. You just don't know it, because the creators producing it have mostly signed NDAs (non-disclosure agreements).

"Brands will have you sign an NDA forbidding you from mentioning they used AI," Mansbridge said. She calls the move "plausible deniability." Consumer trust is still being built, and no one dares go public.

This Trend Started Three Years Ago

Mansbridge said the explosion of "paid UGC" — user-generated content that brands pay for but style to look spontaneous — began roughly three years ago.

Influencers back then were clever. They saw brands were desperate for "real ordinary person using the product" content, so they went knocking on doors, shot in that style, and charged for it. Brands ate it up.

"The authenticity of UGC was never about who made it — it was always about resonance," Mansbridge said. "If the content nails how consumers genuinely feel about a product, it connects. AI just gave brands a smarter, more scalable way to do it."

Sounds perfectly reasonable.

But be careful with the words "perfectly reasonable."

Now It's Even Coming to You

Recently, an artist in Leeds who makes art prints — Zac Rossiter — received an email. A marketing company offered to make him an ad film for free: pick a product, and they'd use their in-house AI studio to generate one for him.

What kind of format? An unboxing video.

The kind where a customer tears open the packaging and reacts with surprise to the camera. It's the most popular format in UGC, because it looks the most like a "genuine recommendation."

Rossiter turned them down.

"I would never work with a company that uses fake AI unboxing videos in place of real people," he said.

I wanted to applaud him.

Back to That Bride Video

I reopened the video. The bride is still crying. The app's name is still on screen. People in the comments are still moved.

There's nothing wrong with AI itself, nothing wrong with saving money, and certainly nothing wrong with running experiments. The real problem is that when a machine pretends to be an ordinary person like you or me making a "genuine recommendation," and you and I can't tell the difference — that chain of trust snaps.

I've never forgotten that line from Mansbridge: consumer trust is still being built.

Yes. Still being built.

Which is exactly why everyone's hiding it.

I don't have the answer. But this is worth thinking about. Next time a piece of "authentic user experience" moves you, take one extra look at that hand — and count the fingers.