That Bride You Scrolled Past, Crying About a Product? She Might Not Exist

Recently I was scrolling through short videos when I came across a bride, tears streaming down her face, telling the camera: "Everyone said the wedding should be phone-free, so I put a disposable camera on every table — and I used this app called Once."

Very real. Very moving.

Then I scrolled down a bit and found another wedding video where the bride was saying the exact same thing, promoting the exact same app. Different person, different filter, nearly identical script.

Something's off here.

Later, a deepfake-detection firm called Reality Defenders ran a round of testing, and the results came back: these brides' faces were AI-generated. The wedding photography app Once was using "real customers" on Instagram who, in all likelihood, didn't exist at all.

I reached out to Once to ask what was going on.

No response.

A New Profession: The "AI Influencer" Producer

Let me tell you a story.

There's a woman in London named Clarissa Mansbridge. She used to be an agent for celebrities like Katie Price. Now she doesn't take on celebrities anymore — she takes on projects, specifically making AI personas for brands.

She has an alter ego called Mia Metaverse. A brand sends over a brief (a creative brief — the project specification): they need a woman in her early twenties, applying sunscreen by a pool in Bali, with their skincare product placed right beside her. Clarissa whips up an AI persona, delivers the video, and the brand posts it to their own account. You couldn't tell it apart from genuine UGC (user-generated content).

She told a journalist something I've read over and over:

Roughly 40% to 60% of the content major brands put out online is AI-made. But many of the creators have signed NDAs, so they can't talk about it.

She calls this "plausible deniability."

Brands are using these personas while simultaneously terrified of consumers finding out. So they make the producers sign non-disclosure agreements — they'll let you create the personas, but they won't let you admit you created them.

Why Are Brands Doing This? Let's Do the Math

Think about it.

How much does a traditional commercial shoot cost?

Clarissa's numbers: $20,000 to $70,000. Photographer, model, location, retouching, reshoots — all of it.

An AI influencer? The same kind of footage — generate a persona, run a few rounds of prompts — maybe a few hundred, maybe a few thousand dollars.

An order of magnitude difference.

And real influencers can "cause trouble" — saying the wrong thing, getting swept up in controversy, raising their rates at the last minute, attracting paparazzi attention. AI personas don't. They'll shoot for you at 3 AM, forever emotionally stable, never charging overtime.

Clarissa said something that cuts deep: "Real influencers killed their own market."

A Statistic That Stopped Me Cold

The consumer organization Which? ran a test — they showed participants a batch of videos and asked them to tell which were real and which were fake.

70% of people couldn't get them all right.

In other words, out of ten people who see this kind of content, seven will be fooled. They think they're watching a genuine customer giving a heartfelt recommendation. What they're actually watching is a face generated by a prompt.

Lisa Barber, Which?'s tech editor, put it bluntly:

Consumers can no longer trust what they see. Companies must be transparent about content made with AI.

So, Does the Law Step In?

This is where it gets interesting.

I dug through the rules of the UK's Advertising Standards Authority (ASA). A spokesperson told me:

Our rules do not prohibit brands from publishing AI-generated advertising content, nor do they require AI to be labeled.

In other words — in the UK, you can absolutely create a fake bride, have her cry on camera about how amazing your app is, and say nothing about it.

The ASA's only catch-all clause is this: ads must not be "misleading" and must demonstrate "social responsibility." But how is "misleading" determined? The spokesperson themselves said it depends on whether a specific ad creates a misleading impression for consumers — it's not about AI being inherently problematic.

The EU is much tougher on this. The AI Act requires that, starting in August 2026, deepfake images, audio, and videos must be clearly labeled as AI-generated or AI-manipulated. No label? That's illegal.

But this legislation doesn't apply in the UK.

Three Types of Brand Responses I've Observed

The first type: Did it. Got asked. Deleted the photos. A Dubai-based fashion brand called Ashle posted a photo of a female model wearing their clothes at a restaurant. Sharp-eyed viewers noticed the model had one too many fingers. After journalists reached out, Ashle pulled the photo from social media. The brand's explanation: those designs were no longer part of the new collection — that's why they removed them. Not because of AI.

The second type: Did it. Admitted it was an experiment. A property design app called Maket, when asked "why is your AI persona praising your app," gave a remarkably honest answer:

AI influencers are a small-scale way for us to test creative concepts and marketing hooks. This isn't core to our marketing — it's just an experiment to understand what resonates with users.

I read between the lines — what they're really testing is whether AI personas can fool users.

The third type: Agencies proactively pitching the service. An artist in Leeds named Zac Rossiter recently received an email: pick one of your pieces, and we'll use AI to create an unboxing video for you, free of charge.

An unboxing video, if you're not familiar, is the format where a customer opens a product on camera and reacts with surprise and delight. It's now considered the content format that "most closely resembles a genuine recommendation."

Rossiter turned them down. He said: "I will never work with an agency that replaces real people with AI-generated unboxings."

How I See It

To be honest, after reading through all the material, my first reaction was torn.

On one hand, AI personas fill a genuine market need. Brands want to cut costs, they want stability, they want controllability — nothing wrong with that. Technological progress shouldn't be demonized on its own.

On the other hand, the crux of this issue isn't AI. It's deception.

If a brand tells you plainly: "This is our virtual persona — we'd love for you to learn about our product." Fine. That's a marketing creative choice.

If a brand doesn't tell you, and instead makes you believe this is a real girl in Bali, a real bride, a real customer doing an unboxing — that's not creative anymore. That's deception.

Mia Metaverse said something that really resonated:

The authenticity of UGC has never been about "who made it." It's about "whether the consumer truth it reflects is real." If the content reflects the genuine value of the product, it can connect with people. AI is simply a smarter, more scalable tool.

I agree with the first half entirely. But the second half — "more scalable" — is exactly what worries me most. Once deception becomes scalable, the harm becomes scalable too.

Especially when seven out of ten consumers can't tell the difference.

So, What Should Consumers Do?

I don't have a definitive answer.

But I've set a few baselines for myself:

First, whenever I see any "real customer recommendation" video, I assume it might be fake. Especially when the face is impossibly beautiful, the emotions are overwhelmingly intense, and the script flows just a little too smoothly.

Second, look at the hands. At this stage of AI-generated personas, an extra finger or a missing finger is still the easiest tell. That Ashle photo was caught out exactly this way.

Third, check the account history. A real person's account will have years — sometimes over a decade — of ordinary life traces. An AI persona's account typically only posts product-related content. It has no past.

Fourth, and most importantly — trust this instinct: when something seems too good to be true, it usually is.

One last thing.

The EU will start mandating AI content labels in August. I'm fully behind this. Not because I'm anti-AI, but because I believe this: transparency is the bare minimum for trust. Once a brand won't even tell consumers whether it used AI, all that's left between the brand and the consumer is suspicion.

And suspicion is the most expensive cost in any deal.