The next time you see a woman on Instagram holding up a skincare product, smiling genuinely, telling you it changed her morning routine, take a second look. There is a reasonable chance she does not exist.

A Guardian investigation published today found that brands are quietly using AI-generated influencers to promote products on social media, deploying synthetic people who look like real customers without any clear disclosure that the content is fabricated. The people commissioning this content are being asked to sign NDAs. The brands involved, when caught, either said nothing or removed the posts.

This is not a niche experiment happening at the fringes of digital marketing. One creator who builds these AI influencers for brands estimates that 40 to 60 percent of influencer-style content from some major brands is already AI-generated.

The specific cases the Guardian investigated give a clear picture of how this works in practice.

A photo app called Once ran multiple Instagram videos showing what appeared to be a bride, in tears, saying she used the app at her wedding. The videos had the texture and framing of genuine user-generated content. They were not. Analysis by Reality Defenders, a cybersecurity firm that specializes in deep fake detection, found the person in the video was almost certainly AI-generated. Once did not respond to a request for comment.

Another example involved Market, an app for designing and planning housing projects. A synthetic woman appeared in their promotional content saying, "I could kiss the interior designer who showed me this." Market, to its credit, did at least respond. The company described it as a small-scale experiment to test creative concepts, not a core strategy. That is a more honest answer than most, but it does not change what viewers saw when the post appeared in their feed.

A Dubai-based fashion brand called Ashle posted a photograph of a woman wearing its clothes at a restaurant. The woman had an extra finger, a classic tell of AI-generated images. When the Guardian reached out for comment, the brand deleted the photographs. The explanation offered was that the designs were no longer part of the collection.

The creator at the centre of several of these campaigns, a former celebrity manager named Clarissa Mans bridge, was direct about the commercial logic. Traditional photoshoots cost between $20,000 and $70,000. AI-generated content costs a fraction of that and comes without the complications of human talent: no bad press, no personal opinions, no scheduling conflicts.

She also explained why this stays hidden. Brands routinely require content creators to sign NDAs preventing them from disclosing that AI was used. "I call it plausible deniability," she said.

When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.

89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.

Start flowing free

The Advertising Standards Authority in the UK was asked directly whether brands are required to disclose when a promotional post features an AI-generated person rather than a real customer. The answer was clear: there is no rule requiring them to do so.

The ASA's position is that it evaluates whether an ad is misleading in its claims about a product, not whether the person delivering the message is real or synthetic.

The EU has taken a firmer line. From August 2026, rules under the Artificial Intelligence Act will require AI-generated or manipulated content, including deep fakes, to be clearly labelled. That legislation does not apply in the UK.

The consumer group Which? put the problem plainly. Its own research found that 70 percent of people cannot reliably tell the difference between real and AI-generated video content. That number makes the NDA strategy look particularly calculated. Brands can use synthetic people, require silence from the people building them, and count on most viewers not knowing what they are looking at.

For people who spend time researching products before buying, this matters more than it might first seem.

User-generated content, which includes unboxing videos, review posts, and "look what I found" style promotions, has become one of the most relied upon signals in purchase decisions. Its entire value rests on the assumption that a real person tried the product and is telling you what they thought.

An AI-generated unboxing video, commissioned by the brand selling the product, is the structural opposite of that. It is a paid advertisement wearing the clothes of peer recommendation. Mans bridge described the logic behind UGC this way: "Brands want content that looks like real, everyday people using their products." That is the tell. The point is to look authentic, not to be authentic.

The Guardian investigation also found that at least one marketing agency is actively cold-pitching small businesses and independent creators with offers to make AI unboxing videos for their products. Leeds-based artist Zac Rossiter received one such pitch and turned it down. Not every business will.

None of this means every review post you see is fake. But it does mean the signals you used to trust have shifted.

A few practical adjustments help:

These are not foolproof. The technology is improving quickly. But they raise the bar for what it takes to mislead you, which is ultimately the only practical defense available until regulation catches up.

Mans bridge made one comment that deserves to sit with you. "Consumer trust is still being built," she said, explaining why brands demand NDAs from the people creating their synthetic content.

That framing is revealing. It treats consumer trust as a resource to be managed quietly rather than something earned through honesty. The brands using AI influencers without disclosure are not just bending a rule that does not yet exist. They are making a deliberate calculation that most people will not notice, and that the ones who do will not be able to prove anything.

That might work in the short term. It tends not to work once the practice becomes widely known, which, given today's investigation, it now is.