You can clone a face. You can't clone trust.

A few weeks ago, Emily Schuman, a lifestyle creator with over 500,000 followers, found herself in an Instagram ad she never shot. She was smiling, holding up a medication from a brand she’d never heard of, and looked perfectly at ease. There was just one problem: it wasn’t her, but a digital clone of her, built from an old selfie posted years earlier and put to work without her knowing.

Two more followed in the weeks after: a make-up ad and a post for a blood-testing company, both from brands Schuman had never had any commercial relationship with.

The story was reported in a Guardian investigation by Arielle Pardes. And striking as Schuman’s case is, the point is that it’s far from isolated: it’s a symptom of something already changing the rules of influencer marketing.

The numbers, once you go looking, are chilling. According to research by Surfshark, cited in the Guardian’s investigation, consumers worldwide may have lost up to $3.7 billion to deepfake scams in 2026, nearly half of it from fake identities on social media.

A McAfee survey of 8,600 people, reported by WhatsTrending, found that 72% of Americans have already come across a fake celebrity or influencer endorsement online. For one in ten, that encounter came at a real cost, in money or stolen personal data.

Meanwhile, “virtual” influencers, the kind generated entirely by AI with no real person behind them, already represent a $1.37 billion business, roughly 4% of all global advertising spend on creators.

Between a digital face that’s openly synthetic and a real face stolen without permission, the line is getting thinner by the day. And for anyone working every day to build credible relationships between a brand and a real person, neither direction is good news.

Among all the stories told in the piece, though, one deserves to be singled out.

Molly Tranchin, a fashion creator with half a million followers, had signed a contract with underwear brand Eby to shoot three promotional videos. As often happens, the agreement included a clause granting the company the right to “reasonable edits” on the delivered content. Eby interpreted that clause its own way: it took Tranchin’s videos and remade them from scratch with AI, in a version very different from the one the creator had actually shot. The two sides will now settle it in court, as Tranchin has decided to sue.

What’s most unsettling about the whole story is that there’s no hacker, no scammer, no fake profile.

There’s one distinction here that matters more than any other: the one between brands that treat authenticity as a problem to manage once it blows up, and those that turn it into a competitive edge.

The first group only notices the problem once reports start coming in, and from there can do little but wait for platforms to step in, which are in no particular hurry to. Not by accident, either. A Reuters investigation estimates that Meta made roughly $16 billion from ads tied to scams or banned products. Hard to expect much urgency from a company that, in the end, profits handsomely from the whole situation.

The second group understood, ahead of everyone else, that trust today can be told as a story. Dove built one of Cannes’ most-awarded campaigns around genuine Reddit reviews of its own product, negative ones included. Aerie made an explicit rejection of AI-generated content part of its brand identity, launching a dedicated creator programme that blew past a six-month goal in a matter of weeks (both cases are covered by Marketing Dive). Even Reddit, as a platform, invested in an entire campaign, “People Are The Best,” built on a simple idea: the one thing no artificial intelligence will ever replicate is a real person (Social Media Today).

None of these brands has solved the deepfake problem, to be clear. But all of them grasped that, at a moment when online trust is running scarce, whoever defends it openly is already a step ahead of everyone else.

Most creator agreements in circulation today were written before anyone could generate a synthetic face in a few clicks. The old clauses on “editing” and “edits” meant touch-ups, filters, maybe a bit of trimming in the edit suite. Nobody had yet imagined they could come to mean rebuilding a person from scratch.

In practice, for whoever builds these agreements, very little changes in theory and a great deal changes in practice. It needs to be spelled out in black and white that artificial intelligence is not a synonym for “editing,” but its own category, with a genuine final approval right over what actually gets published, not just over what was filmed. The way a creator’s likeness is used by partners needs ongoing monitoring, rather than being discovered only once the damage is done. And before signing with any partner, it’s worth checking what they actually do with AI, not just what they claim to do in their press materials.

There’s a temptation, reading stories like this one, to wait for someone else to solve the problem: a stricter platform, a clearer law. Early signs suggest that would be a mistake. A German court has just ruled that Meta is liable for fraudulent ads placed by third parties on its platforms, with fines of up to €250,000 for every future fake ad. And even Meta’s own Oversight Board, its own internal supervisory body, has called the current rules on harmful deepfakes “fundamentally inadequate.” If even internal governance admits as much, real protection will most likely arrive after the damage, not before. In the meantime, whoever builds contracts and partnerships doesn’t have the luxury of waiting for regulation to sort things out: that safeguard needs to be written in today, not left to a referee who will step in tomorrow.