AI and Virtual Influencers vs Human Creators: Should Brands Use Them?

Short answer: it depends on the purpose — and soon also on new labelling rules

An AI or virtual influencer is a fully computer- or AI-generated persona with no real person behind it — think Lil Miquela, Aitana López or Lu do Magalu — that posts on Instagram, TikTok and other platforms as if it were real. Virtual influencers give the brand total control over the message, are available across unlimited languages and markets at once, and can't create a scandal through their own private behaviour. On the other hand, they can't build the same genuine, personal relationship with an audience that a human creator can, and from 2 August 2026 the EU AI Act (Article 50) adds new labelling duties on top of the existing Danish and EU ad-disclosure rules. Which one makes sense depends on purpose, budget and how much control the brand needs — the answer is rarely either/or.

What is an AI or virtual influencer, exactly?

An AI or virtual influencer is a digital persona — a face, a voice and a personality — built with 3D graphics and/or generative AI, with no real person behind the role. The persona "posts" on social media, takes part in brand collaborations and replies to comments exactly as a human influencer would, but every word, image and post is written, designed and approved by a studio, an agency or the brand itself.

That's not the same as a real creator using AI tools — a photo retouch, an AI-generated background, or a cloned voice for a voiceover, for example. There's still a real person with their own history, opinions and accountability behind that kind of content. It's also not the same as an unauthorised deepfake of a named, real person — that's a separate, serious problem with entirely different legal consequences (typically image rights and personality-rights infringement) and sits outside the scope of this article. It's also not the same as a faceless or anonymous creator account — that's a real person who simply doesn't show their face, not a synthetic persona with no one behind it.

Three examples that show the range from stylised to photorealistic

Lil Miquela was created in 2016 by the US studio Brud and today has around 2.6 million Instagram followers. She has appeared in campaigns for Prada, Calvin Klein, Givenchy and Samsung, and her style is clearly CGI but stylised close to photorealistic.

Aitana López was created in 2023 by the Spanish agency The Clueless and has over 265,000 followers. She's noticeably more photorealistic than Miquela — designed to resemble an ordinary, credible young woman rather than an obviously "digital" character — and earns from a mix of brand deals and paid content under her persona.

Lu do Magalu was created by the Brazilian retailer Magazine Luiza, originally as a digital customer-service assistant, and has since grown into a full-fledged influencer figure with over 6 million followers and collaborations with brands including Samsung and Adidas. Her style is more cartoon-like than the other two — and that range, from clearly stylised to hard-to-distinguish-from-real, turns out to have concrete legal significance, as the EU AI Act section below explains.

There is no single reliable figure for how big the virtual-influencer market actually is — competing research firms published 2026 global market estimates ranging from roughly $8 billion to $16 billion for the same year, with no shared methodology. That kind of spread is exactly why this article doesn't quote an overall market size as a fact.

Virtual/AI influencer vs human creator: compared

Why do some brands use virtual influencers — and why do most not?

Advantages

• Total creative control. There's no off-brief risk, because every word and image is approved internally before it's posted.
• Unlimited availability. The persona can "appear" in multiple campaigns, languages and time zones at once with no booking conflicts.
• No personal scandal risk. A virtual influencer can't get caught up in a personal breakup, a court case or a controversial remark the way a real person can.
• Ownership instead of licensing. Depending on the deal with the agency, the brand often owns the persona and all its content outright, rather than licensing usage rights to a creator's existing content.

Disadvantages

• High set-up cost and specialist skill. Building a credible persona requires 3D/AI production and ongoing maintenance — not a task most marketing teams can handle in-house.
• "Uncanny valley" risk. The closer to photorealistic the persona is, the greater the risk it reads as unsettling or artificial to an audience instead of credible.
• No genuine parasocial relationship. An audience can follow and like a virtual influencer, but the deep, personal trust that drives much of the effect of human creators doesn't form in the same way toward a known construction.
• The brand still carries the real risk. In May 2019, Calvin Klein faced heavy criticism for a campaign in which real-life model Bella Hadid kissed Lil Miquela — the backlash centred mainly on accusations of "queerbaiting," but Miquela's artificial status added to the audience's unease about the campaign, and Calvin Klein ended up apologising publicly. The point isn't that virtual influencers are inherently risky — it's that a virtual persona doesn't remove the brand's own responsibility for the creative choices it's used in.

The rules in 2026: disclosure just got harder to skip

The US: the FTC's Endorsement Guides

The US Federal Trade Commission updated its Endorsement Guides in July 2023 to state explicitly that an endorsement from an AI-generated or virtual persona is deceptive if a reasonable consumer might believe it comes from a real person with genuine opinions and experiences — unless it's made clear that the persona isn't real. The FTC has also tightened what counts as valid disclosure: it has to be unavoidable to the audience, not merely technically present somewhere in the post.

The EU: AI Act Article 50 (from 2 August 2026)

The EU AI Act introduces, from 2 August 2026, two separate transparency obligations worth keeping distinct:

• Marking synthetic content (Article 50(2)): providers of AI systems that generate synthetic audio, image, video or text must ensure the output is marked in a machine-readable format as artificially generated. This duty applies broadly — regardless of how stylised or photorealistic the persona is.
• Deepfake disclosure to the audience (Article 50(4)): the deployer — the brand or agency publishing the content — must additionally disclose to the audience that the content is artificially generated, if it qualifies as a "deepfake" under the Act's definition (Article 3(60)): synthetic image, audio or video content that resembles an existing or plausibly existing person, object, place or event, and would falsely appear authentic to someone viewing it.

This is where the range from cartoonish to photorealistic in the section above becomes concretely legally relevant: a clearly stylised character like Lu do Magalu doesn't necessarily resemble a "plausibly existing" real person, so it sits in a greyer zone under the deepfake definition, while a much more photorealistic persona like Aitana López sits closer to meeting the criterion — even though neither depicts a specific, named real individual. The rule doesn't require an identifiable real person behind the persona; it's enough that the result could plausibly be a real person.

A lighter regime applies to content that's clearly artistic, satirical or fictional — but lighter isn't the same as exempt; a minimum level of visible disclosure is still required. Denmark's own Dansk Erhverv confirms in its own guidance that Danish businesses are covered from 2 August 2026 on the same terms as the rest of the EU — no Danish carve-out or exception beyond the general EU framework has been found.

Ad-disclosure law still applies — whether or not the "creator" is real

The AI Act's labelling requirement is a new, separate duty — it doesn't replace the existing Danish and EU rule that commercial content must be marked as advertising. If a brand pays for a virtual influencer to talk about or show a product, that triggers the same disclosure duty under Markedsføringsloven § 6(4) and the UCPD as it would for a human influencer — see the full breakdown in influencer marketing disclosure rules in Denmark and the EU. An artificial face doesn't remove the commercial intent that has to be disclosed.

We haven't found any specific, named guidance from Forbrugerombudsmanden on virtual or AI-generated influencers yet — the same way there's no specific Danish rule yet for employees' own posts, see employee advocacy vs influencer marketing. Until one exists, the cautious approach is to apply the ordinary disclosure duty in full, plus the AI Act's new requirement on top — not to assume one replaces the other.

Decision framework: virtual influencer, human creator, or both?

IF you need total control over the message and have to reach many languages/markets at once with the same visual identity → a virtual influencer can make sense, provided the budget and technical capacity exist.

IF you're selling on trust, authenticity and a felt personal recommendation → a human creator, typically nano or micro, is the stronger route.

IF the budget can't absorb a multi-year set-up and maintenance cost → human creators are the more realistic starting point, whatever the worked example below shows over the long run.

IF the goal is a long-running, recognisable brand universe (like Lu do Magalu) rather than one-off campaigns → a virtual persona can justify the higher set-up investment, in the same way a long-term ambassador program justifies a longer relationship over one-off campaigns.

IF you're unsure whether you can skip disclosing that the "creator" is artificial → assume you have to disclose it. Both rule sets covered above point the same way.

Worked example: when does a virtual influencer pay for itself?

The figures below are entirely hypothetical, for illustration only — not a real agency or customer deal.

A brand wants a consistent, recurring face for its content. Two routes:

Virtual influencer: a one-time cost of DKK 150,000 for design, the 3D/AI model build and a licence agreement with the agency, plus DKK 20,000/month in ongoing production.

Human creators: 20 nano creators at DKK 1,500 per confirmed post per month — the same illustrative unit price used in nano, micro, macro or mega influencers — for a total of 20 × DKK 1,500 = DKK 30,000/month, with no set-up cost.

The cumulative cost of the virtual influencer is DKK 150,000 + DKK 20,000 × number of months. For the human creators it's DKK 30,000 × number of months. The two lines cross when 150,000 + 20,000m = 30,000m — that is, at m = 15 months, where both routes have cost DKK 450,000 in total. Before month 15, the human-creator model is cheaper; after month 15, the virtual influencer is cheaper, and the gap widens by DKK 10,000 a month after that (for example, DKK 90,000 cheaper after two years).

The conclusion isn't that one model is best — it's that a virtual influencer is only the cheaper route if the brand commits to using it for more than a year, and has a budget that can absorb the set-up cost from day one.

Common mistakes

• Assuming disclosure doesn't apply because "there's no real person." The commercial intent is the same whether the face is real or artificial.
• Assuming a virtual influencer is risk-free. The Calvin Klein example shows the brand's own creative choices can still create a real scandal.
• Underestimating the set-up and maintenance cost. A credible persona needs ongoing production, not just a one-time design.
• Overlooking the AI Act's new labelling requirement from 2 August 2026. It's a separate duty on top of existing ad disclosure, not a replacement for it.
• Expecting the same parasocial relationship as with a human creator. Followers can like a virtual influencer without building the deep trust that drives much of the effect of real creators.
• Forgetting to clarify ownership and usage rights to the persona with the agency. Without a clear agreement, the brand can end up only licensing, not owning, the character it paid to build.

Make Influence's perspective

Make Influence's platform and tracking model is built for collaborations with real, external creators — tracked links, commission and campaign management with people who have their own audience. We don't build or license virtual influencers ourselves, and for most of the brands we work with, a human nano or micro creator is both the cheaper and the faster route to trust and sales. A virtual influencer is a different discipline — closer to VFX and AI production than creator sourcing — and it's worth knowing that before comparing the two on price alone. This is our operational view, not a claim that one is always better than the other.

FAQ

Does a virtual influencer have to disclose that it isn't human?

Yes. The FTC's rules in the US require it explicitly, and the EU AI Act introduces an equivalent disclosure duty from 2 August 2026 for content that resembles a real or plausibly real person. Failing to disclose it is deceptive marketing under both rule sets.

Do Danish and EU ad-disclosure rules still apply even though the "creator" isn't a real person?

Yes. The disclosure duty under Markedsføringsloven § 6(4) and the UCPD depends on whether there's a commercial intent and a benefit — not on whether the sender is human. See influencer marketing disclosure rules in Denmark and the EU.

Is a real creator's AI-edited photo the same as a "virtual influencer"?

No. As long as there's a real person with their own name and accountability behind the post, it's still a human creator using a tool — not a virtual influencer. This article covers fully synthetic personas with no real person behind them.

Can a virtual influencer get "cancelled" the same way a human creator can?

Not in the same way, because the persona can't act independently of the brand. But as the Calvin Klein example shows, the brand's own creative decisions around the persona can still create a real scandal and backlash.

Is there a standard price for building a virtual influencer?

No, and this article deliberately doesn't quote a general price list — only a hypothetical worked example to illustrate the break-even logic. The price depends on the agency, the complexity, and how photorealistic the persona needs to be.

Does Make Influence work with virtual influencers?

No. Make Influence's model is built for collaborations with real, external creators. See Make Influence's perspective above for why that's a different discipline.

Is training an AI model on a real creator's content the same as building a virtual influencer?

No. This article covers a fully synthetic persona with no real person behind it. Using a real influencer's existing content to train an AI model is a different question entirely, governed by copyright's text-and-data-mining rules rather than disclosure law. See Can Brands License Influencer Content for AI Training? for that separate topic.

Is a virtual influencer the same as a faceless or anonymous creator account?

No. A faceless creator is a real person who deliberately doesn't show their face; a virtual influencer has no real person behind it at all. See faceless and anonymous creator accounts for that distinction.