Human influencers still deliver more trust; virtual influencers deliver more control. Neither wins outright, and the peer-reviewed evidence says the product usually settles it: virtual personas perform about as well as people when the thing being sold is functional, and noticeably worse when it is experiential, emotional or bound up with identity.[1][2] Our advice to most clients is a hybrid, with a brand-owned persona handling volume and consistency while human creators supply the credibility that only comes from having actually used the product.

The category is no longer a stunt. Lil Miquela made TIME's 2018 list of the 25 most influential people on the internet, a decent result for someone who does not exist.[3] Imma, billed as Japan's first virtual model, spent three days living in a shop window when IKEA opened its Harajuku store in 2020.[4] Aitana López, a pink-haired Barcelona "model" run by the agency The Clueless, reportedly earns up to €10,000 in a good month.[5] The useful question is what they are good at, and what they cost you when they go wrong.

What You Are Actually Buying

Control, and Its Limits

The case for a virtual influencer nearly always starts with control. Rubén Cruz, who created Aitana López, has said his agency built her after too many projects were "put on hold or cancelled due to problems beyond our control", often the talent's.[5] A persona you own does not renegotiate, post something regrettable at 2am or sign with your competitor.

Control is not the same as safety, though. It moves the risk from the talent to the brief. Calvin Klein's 2019 spot in which Bella Hadid kissed Lil Miquela drew accusations of queerbaiting and a public apology from the brand.[6] In 2025 Miquela announced a leukaemia diagnosis as part of a donor-registry campaign for the nonprofit NMDP, and commenters called the fictional storyline disrespectful to people actually living with cancer.[7][8] Neither misstep was the persona's. Both belonged to the people writing it.

The research adds two wrinkles. A study in the Journal of Advertising found AI influencers can deliver brand benefits similar to human celebrity endorsers, but when one misbehaves, consumers tend to read it as something all AI influencers do. Replacing it with a human celebrity softened the damage; swapping in another AI influencer did not.[9] And a 2026 set of experiments found audiences forgive human-like virtual influencers less readily than anime-style ones.[10] The more convincingly human the persona, the more human the standard it is held to.

Cost Structure

Human influencer spend is rental. You pay per post or campaign, and when the contract ends the audience goes home with the creator. A virtual influencer inverts the model: the money goes up front into character design, visual identity and voice, then into ongoing production and community management, and whatever audience it builds belongs to you.

The catch is time. An owned persona starts at zero followers, which makes it a multi-year commitment rather than a campaign line item. For scale, Euronews reported Aitana López's per-ad rate at just over €1,000, with average monthly earnings around €3,000.[5] Respectable for one account, and every euro accrues to the agency that owns her.

Always-On Production

A virtual influencer does not get sick, book out or need flights, and it can post from Melbourne and Tokyo on the same afternoon. That is a genuine operational advantage, but a 2026 analysis of 116 virtual influencers found posting frequency did little for perceived authority or the sense of closeness followers feel; perceived expertise, credibility, visual appeal and a consistent, believable persona did the work.[11] Volume lowers the cost per post. It does not raise the value of one.

Trust, Authenticity and What the Engagement Data Says

The fair summary of the academic work is that people trust people more, and that the size of the gap depends on what is being sold.

A 2024 University of Florida study in the Journal of Product & Brand Management found virtual influencers were consistently seen as less authentic and therefore less effective as endorsers, comparable to humans for functional products and ineffective for symbolic or experiential ones.[1] A 2024 experiment with 275 US Instagram users in the Journal of Business Research landed nearby: virtual influencers' recommendations were judged more useful, particularly for utilitarian products, but consumers identified more closely with humans.[2]

The most practical recent finding comes from a paper in the Journal of the Academy of Marketing Science in August 2026. Across 9,766 Instagram posts from 28 virtual influencers, posts showing the persona with humans it appeared to be interacting with earned about 21 per cent more engagement.[12] Those social ties made the persona seem more human, and therefore more trustworthy. The effect weakened when the companion was also virtual, and was muted for human influencers, who have that box ticked already.

Be wary of the headline engagement rates passed around this category; they say nothing about whether anyone bought anything. The test that matters is the one we apply to any influencer program in our growth work: does attention turn into the outcome you are paying for?

Audience tolerance also has limits. In January 2025 Meta deleted its own AI-generated character accounts, including a persona called Liv, after users and journalists picked apart their invented backstories.[13] A persona that is candid about what it is can build a following. One that borrows an identity it has no claim to tends to get found out, publicly.

Disclosure: The Rules Apply to Pixels Too

A virtual influencer carries two disclosure obligations: the commercial relationship, and the fact that the persona is synthetic.

Australia

The Australian Consumer Law applies to influencers and to the brands that pay them. In findings published in December 2023, the ACCC reported that 81 per cent of the 118 influencers it reviewed were making posts that raised concerns, most often undisclosed brand relationships or vague tags like "sp" and "spon".[14] It is enforcing, too: in March 2026 PhotobookShop paid $39,600 in penalties after the ACCC issued infringement notices over allegedly misleading influencer reviews, posted without disclosing the free product behind them.[15] Separately, the AANA Code of Ethics requires advertising to be clearly distinguishable as such, and Ad Standards applies it to businesses rather than to individual influencers.[16] With a brand-owned persona there is no third party in the frame. The business is the advertiser and, in effect, the influencer.

The ACCC's own guidance illustrates the other trap: an influencer paid to rave about a resort they never visited is likely to be misleading.[17] A virtual persona gushing about a serum it has never worn sits uncomfortably close to that example.

United States

The FTC's updated Endorsement Guides, announced in June 2023, clarified that their definition of endorsements extends to virtual influencers, and warned that a platform's built-in disclosure tool may not be adequate on its own.[18] The Guides also require an endorser presented as using a product to be a bona fide user of it, and hold advertisers liable for misleading endorsements.[19]

Platforms and the EU

Meta requires people to label photorealistic video and realistic audio that has been digitally created or altered, and says it may penalise those who do not.[20] TikTok requires creators to label realistic AI-generated content,[21] and YouTube requires disclosure of realistic synthetic media while exempting content that is clearly unrealistic, such as animation.[22] A stylised, obviously animated persona may not trigger that rule; a photoreal one is precisely what it describes. In the EU, the AI Act's transparency obligations have applied since 2 August 2026, including a requirement to label deepfakes.[23]

The practical standard is simple. Say what the persona is in its bio, use each platform's AI label, and disclose paid relationships in plain words, exactly as you would for a human creator.

A Decision Framework

Start with the job rather than the novelty. Four questions sort most briefs.

  • Does the message depend on first-hand experience? If the claim is "I tried this and it worked", you need a person who tried it.
  • Is the product bought on function or on feeling? Functional, utilitarian products are where the research gives virtual personas parity. Experiential and identity-driven purchases, which covers most of fashion, food and travel, still favour people.
  • How much content do you need, and for how long? A handful of launch posts is a human creator brief. Always-on presence across several markets, sustained over years, is where an owned persona starts to make economic sense.
  • How much control does the category demand? Reputation-sensitive categories can benefit from a persona whose every word goes through approval, provided you accept those approvals are now entirely your risk.

Most programs we would recommend end up hybrid. The persona becomes an owned channel for explanation, demonstration and consistency, and appears alongside real people often, because that is when it performs best.[12] Human creators carry the testimonials and the first-hand claims. Keep a human option in reserve, too: in the Journal of Advertising research, a human replacement is what softened the brand damage after an AI influencer's misstep.[9]

Pick the Job, Then the Face

Virtual influencers are good at being consistent, available and yours. Human influencers are good at being believed. Most brands need some of both, in proportions set by the product rather than by whichever demo went viral last week.

If you are weighing up a persona, our AI influencer team builds them from character brief to content pipeline, disclosure included, and will tell you plainly when a human creator is the better buy. Start the conversation here.

Frequently asked questions

Are virtual influencers more effective than human influencers?

Not across the board. Peer-reviewed studies published in 2024 found virtual influencers are perceived as less authentic than human ones and are less effective for experiential or identity-driven products, but perform comparably when the product is functional. Their real advantages are control, consistency and ownership rather than raw persuasiveness.

Do AI influencers have to disclose that they are not real?

In practice, yes. Meta, TikTok and YouTube all require disclosure of realistic AI-generated video, the EU AI Act has required deepfakes to be labelled since 2 August 2026, and the US Federal Trade Commission treats virtual influencers as endorsers under its Endorsement Guides. Any paid relationship must also be disclosed exactly as it would be for a human creator.

What are the advertising rules for virtual influencers in Australia?

Australia's general advertising rules apply to them. The Australian Consumer Law prohibits misleading conduct by influencers and by the businesses behind them, and the AANA Code of Ethics, administered by Ad Standards, requires advertising to be clearly distinguishable as such. Because a virtual influencer is usually owned by the brand or its agency, that business carries full responsibility for disclosure and for every claim the persona makes.

When should a brand use a virtual influencer instead of a human one?

A virtual influencer makes the most sense for functional products, always-on content across several markets, and categories where tight message control matters, provided the brand is prepared to build an owned asset over several years. Human creators remain the better choice for testimonials, experiential or fashion-led products and anything that depends on cultural credibility. Many brands get the best result from a hybrid of the two.

Can a virtual influencer endorse a product it has never used?

Only with care. The FTC's Endorsement Guides require an endorser who is presented as using a product to be a genuine user of it, and the ACCC treats paid claims about experiences that never happened as likely to be misleading. A persona can explain, demonstrate or showcase a product, but first-hand claims about results are better left to real people who have actually used it.

References