AI-Generated Influencers: The Transparency Fight

TL;DR

• The Guardian published an investigation today revealing brands are quietly deploying AI-generated influencers on Instagram and TikTok — fake brides crying over wedding apps, fake women raving about interior design tools, fake fashion models with extra fingers.
• Reality Defenders, a deepfake detection firm, confirmed multiple brands are using AI-generated personas with no disclosure. Some creators are under NDA. One insider estimates 40–60% of content from major brands is now AI-generated.
• The UK's Advertising Standards Authority says there are no rules requiring AI disclosure. The EU AI Act's transparency provisions kick in August 2 — but Eurocommerce (representing Amazon, H&M, Inditex, Ikea, and others) sent a letter to the EU tech chief on Thursday asking for AI-generated ads to be exempted from the "deep fake" labelling requirement.
• Which? found 70% of people cannot distinguish real from AI-generated video. The consumer group is calling for mandatory transparency.
• The bottom line: The advertising industry has already crossed the Rubicon on AI-generated personas. The regulatory fight is now about whether consumers will be told.

What Happened

On Sunday, the Guardian published a front-page investigation by consumer affairs correspondent Sarah Marsh documenting how brands are deploying AI-generated influencers on social media platforms — primarily Instagram — to promote products while giving no indication the people featured are not real. ¹

The investigation, supported by analysis from cybersecurity firm Reality Defenders, identified multiple brands using the tactic:

• Once, a photo app for weddings, posted videos of a bride crying with joy about using the app. The bride was AI-generated. Once did not respond to requests for comment.
• Maket, an AI-powered interior design app, used AI-generated women in promotional videos with captions like "I could kiss the interior designer who showed me this." Maket acknowledged the practice, calling it "an experiment to better understand what resonates with audiences."
• Ashle, a Dubai-based fashion brand, posted AI-generated photographs of women wearing its clothes — one image showed a woman with an extra finger. After the Guardian approached the brand, it deleted the images, claiming they were removed because the designs were "no longer part of the collection," not because they were AI-generated.

The Guardian also found that marketing agencies are actively pitching AI-generated "unboxing videos" to small businesses — the kind of content consumers typically view as authentic peer recommendations. One Leeds-based artist, Zac Rossiter, was offered a free AI-generated unboxing video by an agency. He declined.

What It Actually Means

This is not a story about a few brands experimenting. It is a story about the advertising industry's supply chain pivoting to synthetic personas at scale, before any regulatory framework exists to govern it.

Clarissa Mansbridge, a former celebrity manager who now creates AI influencers through her "Mia Metaverse" portfolio, told the Guardian: "I'm going to say about 40% to 60% of the content out there from some of the big brands is actually being made through AI, but a lot of the creators are under NDA." ¹

She calls the NDA practice "plausible deniability." Brands want the cost savings — AI-generated content replaces $20,000–$70,000 traditional photoshoots — but they do not want consumers to know.

The economics are brutal for human creators. Mansbridge was blunt: "Human influencers killed the market for themselves. Brands are moving to AI to cut out issues like bad press, personal opinions, hourly rates and photographers."

The Guardian investigation lands in the same week that Eurocommerce — the European retail association whose members include Amazon, H&M, Inditex (Zara), and Ikea — sent a letter to EU tech chief Henna Virkkunen asking for AI-generated advertisements to be exempted from the EU AI Act's transparency requirements. ²

The EU AI Act, which enters into force on August 2, requires companies to "clearly label" AI-generated or manipulated content "constituting a deep fake." Eurocommerce's director general Christel Delberghe argued that AI-generated ads "not intended to mislead users" — her example was "generating an image of a living room to showcase a sofa" — should not be included under the definition of "deep fake."

The timing is not coincidental. The retail industry knows the transparency rules are coming and is racing to carve out exemptions before they take effect. The argument — that labelling "a very large share of AI-assisted content" would "dilute the value of the disclosure to consumers" — is a remarkable piece of rhetorical jiu-jitsu: we're using AI so much that telling you would be meaningless, so we shouldn't have to tell you.

The Tension Between the Two Stories

Read together, the Guardian investigation and the Eurocommerce letter reveal the actual state of play:

The Eurocommerce letter's framing — that AI-generated ads showing a sofa in a living room are harmless product visualisation — deliberately elides the actual practice the Guardian documented: AI-generated people pretending to be real customers having genuine emotional experiences with products. A fake bride crying over a wedding app is not a product visualisation. It is a fabricated testimonial.

Hype Deconstruction

This story is not about AI being "inherently problematic," as the ASA spokesperson carefully framed it. The ASA's position — that it would only assess whether an ad is "misleading" rather than whether AI use itself is an issue — is a category error. The deception is not in the product claim. The deception is in the persona making the claim.

A real human being who says "I could kiss the interior designer who showed me this" is making a testimonial. An AI-generated persona saying the same thing is making nothing — it has no experience, no opinion, no interior designer it could kiss. The statement is fabricated at the level of the speaker, not the product.

The Which? finding that 70% of people cannot distinguish real from AI-generated video means the "misleading" test the ASA proposes is functionally impossible for consumers to apply. You cannot be misled by something you cannot detect.

Stakeholder Landscape

Who benefits:

• Large brands and retailers: AI-generated content costs a fraction of traditional production. Zalando reported cutting content production costs by 90% using AI. ² The NDA ecosystem protects them from consumer backlash while they scale.
• AI content creation platforms and agencies: The "Mia Metaverse" portfolio and similar services are building businesses on the gap between AI capability and consumer awareness.
• Platforms (Meta/Instagram, TikTok): AI-generated content drives engagement and ad revenue. Neither platform has announced specific disclosure requirements for AI-generated influencer content.

Who loses:

• Human influencers and content creators: The $20,000–$70,000 photoshoot is being replaced by a prompt. Mansbridge's assessment that "human influencers killed the market for themselves" is candid but incomplete — the market was killed by economics, not behaviour.
• Consumers: Deprived of the ability to distinguish genuine customer experiences from synthetic ones. The entire premise of user-generated content as a trust signal collapses when the "user" does not exist.
• Small businesses: Pitched AI-generated "authentic" content by agencies, they face a choice between participating in deception or competing at a disadvantage against those who do.

Who is affected but not yet engaged:

• UK consumers: The EU AI Act's transparency provisions will not apply in the UK. The ASA has no AI-specific disclosure rules. British consumers are in a regulatory gap.
• Children and vulnerable consumers: The Guardian investigation did not specifically examine content targeting minors, but the combination of undisclosed AI personas and platforms with large youth audiences (TikTok, Instagram) creates obvious risks.

Cross-Layer Implications

Regulatory: The Eurocommerce exemption request is a live political fight. If the EU grants it — even partially — it establishes a precedent that AI-generated personas in advertising are not "deep fakes" so long as the product being advertised is real. That would hollow out the AI Act's transparency provisions for the largest category of AI-generated content consumers encounter daily.

Commercial: The NDA ecosystem documented by the Guardian suggests that brands are aware of the reputational risk and are managing it through secrecy rather than disclosure. This is fragile. A single high-profile expose of a major brand using AI-generated influencers to promote products with safety or health implications (skincare, supplements, financial products) could trigger a consumer trust crisis that makes voluntary disclosure look cheap by comparison.

Platform: Meta and TikTok have AI content labelling tools, but they are opt-in and inconsistently applied. Neither platform has announced mandatory labelling for AI-generated influencer content. The Guardian investigation effectively demonstrates that voluntary disclosure is not working.

Talent market: The human influencer economy — estimated at $21 billion globally — is being undercut from below by AI-generated content that is cheaper, faster, and never has a scandal. The NDA practice means many human creators are complicit in their own displacement, unable to discuss the work they are doing to train or compete with AI systems.

What This Means for You

If you are a consumer: There is currently no reliable way to distinguish AI-generated influencer content from real human testimonials on social media. The Which? finding — 70% failure rate at detection — means your own judgement is not a reliable filter. The most practical defence is source awareness: content from brands you do not already trust should be treated as potentially synthetic regardless of how "authentic" it appears.

If you are a marketer or brand manager: The NDA approach is a short-term cost play with long-term reputation risk. The Guardian investigation demonstrates that AI-generated content is detectable — Reality Defenders identified it, and the extra-finger problem is not going away. If your agency or content team is proposing AI-generated influencer content, ask whether you would be comfortable with the disclosure being made public. If the answer is no, the practice is not ready for deployment.

If you are a content creator: The economics Mansbridge describes are real and accelerating. The differentiation strategy is not to compete on cost — AI will always win that race — but on verifiable humanity. Creators who can demonstrate they are real people with real experiences will command a premium as synthetic content floods the market. The NDA practice is a trap: it pays bills today but erodes the value of the entire category.

If you are a regulator or policy-maker: The Eurocommerce letter is a test of the EU AI Act's coherence. If AI-generated ads are exempted from transparency requirements, the Act's "deep fake" provisions will apply to political disinformation and non-consensual imagery but not to the AI-generated content consumers encounter most frequently. That is a policy outcome worth making explicitly, not achieving through industry lobbying.

Uncertainty Ledger

• What we don't know: The full scale of AI-generated influencer deployment. Mansbridge's 40–60% estimate is from an industry insider with a commercial interest in the practice. Independent measurement does not exist.
• What would change the analysis: If Meta or TikTok announced mandatory AI-content labelling for sponsored posts, the NDA ecosystem would collapse overnight — brands could not claim "plausible deniability" if the platform itself applied the label.
• What to watch: The European Commission's response to the Eurocommerce letter. A decision is expected before the August 2 enforcement date. Also: whether the ASA faces pressure to introduce AI-specific disclosure rules following the Guardian investigation. The ASA's current position — "nothing in our rules that prohibits this" — is a statement of fact that may become politically unsustainable.

Bottom Line

Brands are deploying AI-generated influencers at scale, under NDA, with no disclosure to consumers. The UK regulator has no rules against it. The EU's rules are weeks from taking effect, and the retail industry is already lobbying to be exempted. The 70% of consumers who cannot tell the difference are the product — their trust is being harvested before they know the game has changed. The window for establishing transparency norms is closing, and the industry is spending it asking for permission to keep the lights off.

Sources: