AI influencers and virtual characters: influencer marketing when the talent does not exist

The model has freckles, a blunt bob, and a few million followers. She posts from Los Angeles, releases music, and has fronted campaigns for luxury fashion houses and electronics brands. She also does not exist. Her images are rendered by a studio, her opinions are written by a team, and her "life" is a script. Welcome to the strange, fast-growing corner of influencer marketing where AI influencers and virtual characters are landing brand deals, where some of the most reliable partners on social media are people who were never born.

Marketing Dive's coverage of a PitchBook report notes that brands like Dior, Calvin Klein, and BMW have all worked with one of the most famous of these synthetic personas, a CGI character who sits near three million followers on a single major social network. That character's developer was last valued at $144.5 million before an acquisition. Fake people, it turns out, generate very real money.

So what actually happens when a brand hires someone who does not exist? That question now sits at the center of a genuine industry fight. Advocates of AI influencers point to lower costs, total creative control, and zero scandal risk. Critics point to collapsing trust, regulatory exposure, and a growing audience backlash. This article walks through both sides with real numbers, real campaign examples, and the research that explains why synthetic personas can both work and backfire.

What AI influencers and virtual characters actually are

The terminology gets messy, so let's pin it down. A virtual influencer is a digital character created with software, given a personality and backstory by a team, and operated on social media as if it were a real creator. Some are stylized and obviously animated. Others are photorealistic enough that followers argue in the comments about whether they are human. And a newer wave of AI influencers goes further: the character's images, and sometimes its captions and replies, are produced largely by generative AI systems.

There is also a growing bookshelf behind this shift. Practitioner handbooks like Sevil Yesiloglu and Joydeep Sharma's influencer marketing text and Gordon Glenister's strategy handbook documented how brands industrialized creator partnerships, and Brittany Hennessy's guide for creators codified the deal norms synthetic personas now disrupt. The persuasion mechanics underneath were mapped decades ago in Robert Cialdini's Influence, and Superfandom by Zoe Fraade-Blanar and Aaron Glucksmann explains the audience psychology that makes character-based marketing work at all. For brand-side context, Joel Backaler's Digital Influence covers cross-border creator campaigns, and an open access multidisciplinary book on the dynamics of influencer marketing compiles the current academic view, a literature the Library of Congress research guide now tracks alongside the early practitioner classic by Duncan Brown.

Researchers writing in Fashion and Textiles traced how the number of virtual influencers grew from 27 in 2017 to 125 by 2019, and how brands from Samsung to Balmain to Louis Vuitton began partnering with them. A more recent mapping review in Frontiers in Communication analyzed 116 peer-reviewed articles on virtual influencers published between 2015 and 2025, showing the topic has moved from curiosity to a serious research field covering credibility, engagement, and consumer psychology. A systematic review of virtual influencer attributes and purchase intention in Frontiers in Human Dynamics consolidates that evidence further, and industry-facing analysis like Emre Danisan's AI influencer guide and Storika's guide to multi-market creator campaigns translates it into practice.

The scale is no longer trivial. A roundup of virtual influencer statistics compiled by AutoFaceless pulls together industry figures suggesting the virtual influencer market reached roughly $8.3 billion in 2025, with projections toward well over a hundred billion by the early 2030s at the optimistic end. For the broader backdrop, GRIN's statistics collection, the annual benchmark report, and its companion statistics roundup track total creator marketing spend, while HubSpot's state of marketing research documents the budget shift toward creator partnerships that synthetic personas are competing for. Even if you haircut those numbers aggressively, the direction is unambiguous: synthetic personas are attracting real budgets.

The cousin that proves audiences do not care who is real

There is also a related but distinct phenomenon worth naming: performers who use animated avatars while streaming and creating content, with a real human underneath providing voice and personality. A documentary on virtual entertainers profiled one talent agency of nearly 90 such virtual performers whose channels collectively exceeded 97 million subscribers, with the parent company booking annual revenue in the billions of yen and listing on the Japanese stock exchange. The performers behind those avatars are human. The characters the audience bonds with are not. That distinction matters later, because it shows audiences will happily form deep emotional attachments to characters they know are synthetic, as long as a real personality animates them. Video explainers have amplified the trend: a business news panel on AI influencers landing brand deals debated whether they will replace human creators entirely, BBC coverage of the trend focused on the ethics concerns, a well-known media entrepreneur argued synthetic personas will make companies millions, and a marketing explainer on parasocial relationships walks through the psychological mechanism that makes any of this work.

Why influencer marketing teams hire people who do not exist

Ask a marketing team why they would sign a contract with a character and the answer is almost always control. A virtual persona never cancels a shoot, never raises its rate mid-campaign, never gets photographed at 2 a.m. doing something regrettable, and never dies. The company that owns the character owns the schedule, the message, the look, and the archive. That is a level of control no human talent contract can deliver.

The economics reinforce the control argument. According to Euronews' reporting on a Spanish AI model, the Barcelona agency behind one pink-haired synthetic character built her precisely because real influencers kept derailing campaigns: projects were paused or cancelled because of the human talent's availability, costs, or demands, not the creative work. The synthetic model now earns up to 10,000 euros a month for her agency from ads, brand deals, and subscription content, at a marginal production cost of a design team's time. Fortune's interview with the same agency captures the motivation bluntly: they wanted to stop depending on people with egos and manias.

The demand is not hypothetical. That same agency told reporters it was inundated with requests from brands wanting their own personalized synthetic model, because a character cannot be fired by its own scandal and cannot quit. Coverage by Entrepreneur and Cut The SaaS traced the same pattern: brands want a controllable, scandal-proof face, and a behind-the-scenes documentary on her creation shows the full team of creative directors, writers, and strategists behind the character, with her reference history documented in encyclopedia entries on the best-known virtual influencer. And industry benchmarks aggregated by Grey Journal suggest AI influencer campaigns can run 30 to 50 percent cheaper than comparable human campaigns once logistics, reshoots, and talent management are factored in, with reported engagement rates sometimes running higher than the human baseline.

The long game: when a character becomes a brand asset

One Brazilian retail giant is the textbook case for long-term thinking. Its virtual spokeswoman has been active since the early 2000s, has tens of millions of followers across platforms, and is reportedly the most followed virtual character in the world, as documented in Vinfluencer's campaign retrospective. The lesson from that two-decade run: the biggest returns come from treating a virtual character as a compounding brand asset, not a one-shot stunt.

The trust problem at the center of it all

Here is where the story gets complicated. Influencer marketing works, when it works, because of trust and perceived authenticity, a mechanism documented across industry research from Sprout Social's benchmark statistics and Meltwater's engagement data. People buy from creators they feel they know. A synthetic persona engineered by a marketing team is, by construction, the opposite of authentic, and consumers appear to know it, a gap Sociallyin's statistics roundup and Thunderbit's benchmarks both surface when tracking how audiences report trusting branded creator content.

Researchers at the University of Florida ran a study of 352 social media users comparing human and virtual influencers across product types, summarized in the university's Pixels vs. People research summary. They found virtual influencers were consistently perceived as less authentic than humans, and that this deficit translated into less favorable brand attitudes and lower purchase intentions. Other work complicates the picture: a meta-analytic review in the Journal of the Academy of Marketing Science synthesizes results across dozens of studies, and a source-credibility study published through Redalyc found that consumer skepticism plays a decisive mediating role in whether virtual influencers can move purchase intention at all.

The psychology runs deeper than simple distrust. Parasocial relationships, the one-sided bonds audiences form with media figures, are the engine of influencer persuasion, and they do form with synthetic personas. A study published in the Humanities and Social Sciences Communications journal documents how beauty creators' credibility attributes and parasocial interaction drive purchase intention, and research on anthropomorphized virtual influencers in the Journal of Consumer Marketing shows those bonds do transfer to virtual characters. Work on anthropomorphism in virtual influencer marketing in the Journal of Retailing and Consumer Services found that humanlike traits in moral virtue and cognitive experience boost both credibility and parasocial bonding, with survey evidence from 430 respondents. A cross-case academic analysis by Khalfallah and Keller comparing five of the most popular virtual influencers found that highly realistic characters create intense parasocial interaction while suffering authenticity problems, whereas more stylized characters sidestep some of that backlash. The audience knows it is watching a fiction, and a fiction that pretends otherwise gets punished.

Who gets synthesized: the representation problem

There is also an uncomfortable pattern in who these characters are. An analysis in The Conversation highlighted research from a University of Toronto doctoral candidate noting that many virtual influencers are crafted as young women of colour, often created by teams that do not share those identities. When a brand hires a synthetic persona, it inherits that critique along with the character.

And then there is the comfort gap. Statistics compiled by AutoFaceless from multiple industry surveys suggest that while a majority of consumers follow at least one virtual influencer, roughly 46 percent remain uncomfortable with brands using them. A viral thread discussion on AI-created virtual influencers stealing business from humans captures the mood in the wild: the most upvoted comment argues that influencers already carried associations of being fake and superficial, which is exactly why replacing them with literal fakes felt, to many readers, like no great loss. That is a brutal read on the influencer category's own trust deficit, and it cuts both ways for brands.

When the face looks almost human: the uncanny valley problem

Robotics researchers noticed decades ago that people respond warmly to clearly non-human characters, but that warmth collapses when a figure looks almost but not quite human. That discomfort zone is called the uncanny valley, and it is a live commercial risk for AI influencers chasing photorealism.

An experimental study in the Journal of Business Analytics tested AI versus hybrid versus human endorsers across anthropomorphism levels and found an inverted U-shaped effect: brand attitude rises as a character becomes more humanlike, peaks at a moderate level, then falls sharply as the character approaches full realism without delivering full human behavior. The same paper found that high technological transparency about the AI nature of the endorser attenuated the uncanny valley effect by roughly 60 percent. Honesty, in other words, measurably reduces the creep factor.

Research published in the Psychology and Marketing journal found consumers report uncanniness when comparing highly humanized virtual influencers with less humanized ones, and that certain social cues in the content can alleviate the discomfort. Meanwhile a study on virtual live streaming published through the U.S. National Library of Medicine showed that when virtual presenters appear almost, but not perfectly, human, they evoke discomfort and erode perceived affinity. Practitioners have translated this research into audit checklists: Dream Farm Agency's guide to the uncanny valley effect lists the concrete visual and behavioral triggers that make synthetic characters feel creepy, from dead eyes to over-restricted movement.

The strategic takeaway is counterintuitive. A visibly stylized character often outperforms a nearly photorealistic one, because the audience is never forced to resolve the question of whether it is being deceived. The best-performing synthetic personas tend to be either unmistakably animated or unmistakably disclosed.

Do AI influencers actually perform: the numbers

Performance claims in this space deserve skepticism, because the hype cycle rewards big numbers. Still, a consistent pattern has emerged across several independent sources.

Engagement is where synthetic personas genuinely shine. Stagwell's analysis of the virtual influencer wave reports that virtual influencers have delivered roughly three times the engagement rate of traditional influencers in brand campaigns, attributing part of the gap to novelty and visual consistency. Benchmark roundups like AutoFaceless' statistics review cite cases of engagement up to ten times higher for professional-grade virtual personalities versus campaigns without them.

Conversion is where the story sours. A widely cited Digiday analysis of campaign data shared by a creator marketplace platform found that 86 percent of brands consented to include AI creators in their campaigns in October 2024, and that by late 2025 the share had fallen to 60 percent. The platform's co-founder attributed the drop to brands' growing wariness of backlash against AI usage. Reach and attention are easy to earn with a synthetic persona. Trust-driven sales are not.

That matches the qualitative signals from practitioner communities. In a marketing automation community discussion on whether AI influencers are changing brand content strategies, practitioners acknowledged the flexibility and control gains while flagging that the trust factor usually tanks performance. And in a community thread on AI-generated creators going viral, one brand owner claimed to have cut costs by 5,000 dollars a month since switching to AI-generated content, while another commenter offered the bleakest possible read: a decent number of those followers are bot accounts, meaning AI is performing for AI.

Attention versus trust: the performance split

Research on the relative efficiency of the two formats is still young: a study on the disruption of influencer marketing comparing AI-generated virtual influencers against humans, published in the Journal of Information and Knowledge Management, finds efficiency gains concentrated in cost and consistency rather than persuasion, and human-computer interaction research presented at the CHI conference documents measurable hesitancy in how willing users are to engage with AI-generated influencer content. The honest synthesis: AI influencers reliably outperform on attention metrics, sometimes dramatically, and unreliably on trust metrics, sometimes catastrophically. Which of those two matters more depends entirely on what the campaign is for. Awareness plays favor the synthetic option. Consideration and conversion plays favor humans, or at least demand a very well-executed hybrid.

Brand campaigns with synthetic personas: what worked and what did not

The case history now runs long enough to draw patterns. Vinfluencer's campaign retrospective documents a furniture brand that placed a virtual character's images in physical store windows and on in-store screens, turning the campaign into a real-world destination and driving foot traffic and viral coverage, a template later adapted by luxury houses. The same retrospective details a tech brand's smartphone campaign with a synthetic persona used explicitly to reach younger, digitally native audiences less responsive to traditional celebrity endorsements.

Sozee's roundup of brand case studies covers campaigns from Prada, a Chinese market activation fronted by a locally resonant AI persona, a fast food chain's virtual founder, and sportswear and cosmetics brands, reporting strong impressions and cost-per-engagement gains across the set, while carefully noting that results depend on visual consistency, clear disclosure, and platform-native formats.

Fashion offers the most instructive failure-and-recovery arcs. A famous luxury house introduced a squad of virtual models and faced immediate accusations of sidelining Black models, a backlash covered extensively across trade press and still cited in academic reviews of virtual influencer marketing, including the Emerald book chapter on virtual influencer marketing and the metaverse and published research on virtual influencer marketing strategies. The most comprehensive profile of the best-known CGI character's history, including a 2018 sabotage storyline that alienated fans and forced an apology, is documented in Ars Technica's investigation, and character profiles for the major virtual humans are maintained at the Virtual Humans directory, including the Tokyo-based virtual model who fronted the hybrid campaign above. Earlier collaborations, like the retail back-to-school campaign documented by Marketing Dive, showed the format could drive store traffic, and brand-side commentary like Stack Influence's assessment of whether virtual influencers are opportunity or gimmick captures the current indecision, alongside Sprout Social's explainer on virtual influencers and Dream Farm Agency's localization analysis. The lesson was not that virtual models fail, but that synthetic diversity is a minefield, as NC State's fashion business research on virtual brand faces makes clear in cataloging both the creative freedom and the reputational exposure: a brand that replaces underrepresented humans with synthetic approximations of them will be judged as if it had fired people, because in the audience's mind, it effectively did.

The smartest play many brands have landed on is the hybrid. Marketing Dive reported on a fashion house campaign that paired a virtual influencer with real celebrity ambassadors across a video series exploring identity and self-expression, explicitly using the synthetic character to underline the campaign's theme about the changing definition of what is real. When the artificiality is the message rather than a hidden cost, the trust deficit transforms into a talking point. Case study collections from Sprout Social, Hootsuite, Later, and Bazaarvoice provide broader context on how these campaigns sit inside the influencer ecosystem, foundational guides like Hootsuite's ultimate guide still anchor most brand playbooks, and platform-oriented guides like Shopify's influencer marketing guide document the payment structures that synthetic personas quietly upend. Teams running AI-assisted creator programs are also building internal guardrails: Hootsuite's experiment with an AI social strategy documents the workflow reality, and their AI compliance guide shows how marketing teams now audit synthetic content before it ships.

The regulation wave: disclosure rules arrive on three continents

For years, the rulebook on AI influencers was mostly empty. That era is over. Three regulatory moves now define the compliance landscape for any brand working with synthetic personas.

First, the United States. The Federal Trade Commission's Disclosures 101 guidance for social media influencers already requires that commercial relationships be disclosed clearly, and the Commission's revised Endorsement Guides updated the definition of an endorser to explicitly cover fabricated and virtual endorsers, language multiple commenters asked the agency to make explicit. The FTC's final rule banning fake reviews and testimonials went further, prohibiting testimonials that misrepresent the endorser's existence or experience, with civil penalties available. And per analysis by Coblentz Law on AI influencer liability, a New York statute enacted in late 2025 requires advertisements using a synthetic performer to include clear disclosure, effective mid-2026, while the FTC's own question and answer document on the reviews and testimonials rule confirms that hired influencer posts count as testimonials under the rule, and McDermott's legal analysis pegs potential penalties in excess of 50,000 dollars per violation, and legal commentary in JD Supra notes there is currently no explicit federal rule requiring virtual influencers to disclose they are not human, but an omission that deceives can still be treated as a deceptive practice. Translation: the absence of a rule that says "label your robot" is not a loophole.

Second, the European Union. The AI Act's transparency obligations under Article 50 took effect in August 2026, per the European Commission's official FAQ and its implementation announcement: users must be clearly informed when they are interacting with an AI system rather than a person, including avatars, and deepfake-style synthetic content must be labeled at first exposure. Law firm breakdowns from Orrick and Greenberg Traurig detail what compliance looks like in practice: visible labels, not buried fine print, and disclosures at the start of an interaction, not in end credits.

Indias double disclosure rule and the global direction of travel

Third, India. The Advertising Standards Council of India issued guidelines requiring that virtual influencer content carry a persistent label identifying the persona as AI-generated or virtual, on top of the standard paid-partnership disclosure, as covered in Sansa Legal's analysis of the ASCI guidelines and IndianTelevision.com's report on the draft labelling framework. The watchdog's draft guidelines, available as a published PDF, sort AI usage in ads into risk tiers, with virtual influencers firmly in the mandatory-labelling tier, a framework also summarized by ALG India and Reelax's compliance guide.

The direction everywhere is identical: if your influencer is synthetic, the audience has a right to know, and burying that fact is becoming a legal liability rather than a creative choice. Brands that built virtual personas on ambiguity are now retrofitting disclosure, and legal analyses like the European Parliament's study on influencers and consumer protection make clear that regulators view the entire influencer disclosure regime through a consumer-trust lens, not a marketing-freedom one.

What it means for human creators with real audiences

If you are a human creator, the rise of synthetic personas is not a death sentence, but it is a forced specialization. The exact capabilities AI influencers commoditize, polished visuals, consistent output, perfect availability, are no longer differentiators for anyone, human or machine. What remains scarce is a real relationship with a real audience, demonstrated taste, accountability, and the trust that accrues to a person who has something to lose.

Practitioner communities are converging on this read. In a creator community thread on whether AI will replace influencers, the prevailing answer was no, but with teeth: creators whose only value is appearing on camera holding a product are in danger, and those whose value is personality, perspective, and relatability are not. Another commenter noted the spread of open contempt for generic AI content, with audiences calling it slop in the comments. And in an entrepreneur community discussion about brands quietly replacing real influencers with AI personas, the sharpest observation was that audiences already treat the influencer-follower relationship as partially manufactured, which is precisely what makes a human swap feel low-stakes to brands and unforgivable to fans when discovered.

The academic evidence backs the specialization thesis. A four-experiment study published in Psychology and Marketing found human influencers are more persuasive than virtual ones in prosocial contexts, driven by perceived similarity and post authenticity. Research on AI-generated versus human-created advertising in Equilibrium and on trust and perceived humanness in luxury advertising in the Journal of Retailing and Consumer Services both point the same direction: when trust is the bottleneck, humanness is the advantage. The advertising community discussion on the future of influencer marketing adds the practical layer: AI is already reshaping how brands select partners, with selection tools analyzing engagement, sentiment, and historical performance rather than raw follower counts.

None of this means the human category is safe by default. It means the bar moved. The creators who survive the synthetic wave are the ones whose recommendation carries weight because of who they are, not what they look like.

How to decide if a synthetic persona belongs in your next campaign

Strip away the hype and the decision reduces to a handful of questions worth answering honestly before signing anything.

What is the campaign actually for? Attention and awareness objectives are where synthetic personas earn their keep, because novelty and visual polish are cheap to buy at scale. Anything that requires the audience to believe the recommender, from product endorsements to expert-adjacent claims, inherits the authenticity deficit the research keeps measuring. And anything involving children, health, or financial products sits squarely in the highest-risk zone of every regulatory framework currently in force.

Who owns the character? A brand-owned persona is a depreciating-or-compounding asset that the brand controls forever, which is why agencies report being flooded with requests for bespoke synthetic models, per the Euronews interview. Renting an established third-party virtual influencer works more like a celebrity buy: you inherit the character's existing fanbase and existing controversies, including the cultural critiques attached to how it was designed.

How transparent will you be? The research on disclosure is encouraging: transparency about an endorser's synthetic nature measurably reduced the uncanny valley penalty in experimental settings, as documented in the Journal of Business Analytics study, and a study on disclosing the virtual nature of virtual influencers published in ScienceDirect found that prominent disclosures can reduce advertising effectiveness while subtle disclosures have a less negative impact, raising a real design tension regulators have already resolved in favor of prominence. Given that the EU, the US, and India now all require some form of clear labelling, the honest move is also the legally safe one.

The four questions to answer before you sign a synthetic persona

What happens when the audience finds out, not if? Community sentiment analyzed across marketing forums suggests audiences forgive characters that never pretended to be human and punish brands that let ambiguity do the persuading. The cultural critique in The Conversation is a preview of the coverage a brand gets when its synthetic persona design choices read as exploitation rather than creativity.

Answer those four questions and the decision usually answers itself. If you want help evaluating real human creators with verified audience data instead of gambling on a character, that is exactly the kind of problem platforms like Infmap were built to solve, with public creator profiles, structured deal workflows, and payment handling from negotiation to payout in one place.

The future: coexistence, not replacement

The loudest takes on AI influencers come from the extremes: either synthetic personas will devour the creator economy, or they are a gimmick that will evaporate. The evidence supports neither. What the data shows instead is a split market. Virtual characters are becoming permanent infrastructure for attention, reach, and controlled messaging, with budgets that were once experimental now recurring, and agencies building stable rosters of synthetic talent. Humans retain the trust franchise, the conversion advantage, and the accountability that audiences and regulators both increasingly demand.

Watch three developments closely. The trade press is already tracking the pivot: Adweek's analysis of virtual influencers frames them as a structural shift in how marketers buy attention, GRIN's comparison of influencer marketing versus paid social shows where the budget pressure lands, and Sprout Social's trends research tracks how measurement expectations are tightening across both human and synthetic campaigns, with HypeAuditor's state of the industry report documenting the audit infrastructure brands now demand. Regulation is hardening fast, with the EU's labelling regime now live, New York's synthetic performer disclosure law effective mid-2026, and India enforcing double disclosure for virtual personas, so the era of ambiguity-driven virtual influencer marketing is legally ending. Platform tooling is democratizing character creation, which will flood feeds with mediocre synthetic personas and push the quality bar, and production costs, in both directions. And audience literacy is rising: the generation that grew up with avatars and animated characters does not particularly care whether an entertainer is human, but it has zero tolerance for being lied to about it, as the parasocial research on avatar-based entertainment already suggests.

The brands that will win are the ones that stop asking whether AI influencers are good or bad and start asking what each type of partner is actually for. A synthetic persona is a broadcast instrument with perfect compliance. A human creator is a trust instrument with imperfect compliance. Confuse the two and you pay for it, in credibility or in penalties. Use each for its native strength and the question of who exists stops mattering, because the audience knows exactly what it is buying from whom.

If you are weighing your next campaign and want the human side done right, explore how Infmap structures creator partnerships from discovery through delivery, check out pricing for your role, and get started free at infmap.com/register. For more context on the surrounding industry, read our takes on why people trust influencers more than brands and how brands detect fake engagement.