When the Person in Front of the Camera Doesn't Exist: AI Virtual Influencers, Synthetic Endorsements, and a New Media Literacy Challenge

# When the Person in Front of the Camera Doesn't Exist: AI Virtual Influencers, Synthetic Endorsements, and a New Media Literacy Challenge

The influencer economy excels at turning “I’m like you” into a commercial asset. Breakfast, skincare, anxiety, clothes, studying, insomnia, tidying a room—even vulnerability itself—can all be packaged as content. We used to assume there was at least one premise: the person in front of the camera is real. Now that premise is starting to loosen.

The rapid rise of AI virtual influencers, synthetic endorsers, and generative-content studios offers brands and platforms an almost perfect fantasy: an endorser who is never late, never becomes embroiled in a scandal, never misspeaks, never raises a fee, never needs rest, can speak any language around the clock, and can address many markets at once. Unlike a conventional animated character, this figure need not look obviously nonhuman. It can resemble a real young person, speak warmly, use convincing facial expressions, and maintain a seemingly consistent personality across platforms.

The question is whether existing advertising-disclosure rules are adequate when a person who does not exist begins building trust with viewers as though they were real.

Virtual Characters Are Not New; Passing as Real Is the New Risk

Animated characters, game figures, virtual singers, and brand mascots have existed for a long time. Viewers usually know they're created roles, so they can appreciate them within appropriate frameworks. Generative AI changes the equation by making these figures nearly indistinguishable from real people and allowing them to reproduce the conventions of influencer content: morning selfies, bare-face sharing, breakup confessions, before-and-after comparisons, live interaction, and "I really like this product".

These forms have commercial value because viewers believe there's a real person behind them living, using, choosing, and taking responsibility. When brands reproduce that trust through a synthetic figure without clearly disclosing that the person does not exist, the result is more than a creative image: interpersonal trust becomes a scalable interface.

The issue should not be framed as “all AI content is deceptive.” A clearly labeled virtual character can be a creative brand asset; a synthetic endorser deliberately made to seem like a real user may cross into misleading. The key has never been which model generated the image, but whether ordinary consumers can reasonably understand the content source, commercial relationship, and degree of reality at the moment.

An “AI-Generated” Label Is Not Enough

AI endorsement content involves at least two kinds of disclosure. First is the person or content origin: was the figure AI-generated, a digital double, or a synthetic actor? Second is the commercial relationship: is this paid advertising, brand collaboration, gift exchange, or other interest-based recommendation?

The U.S. Federal Trade Commission's Influencer Disclosure Guide requires that important relationships between influencers and brands be clear, easy to see, and placed with the recommended content—not hidden in bios, video descriptions, or a long string of tags.

But "AI-generated" doesn't tell viewers whether this is an ad; "brand collaboration" doesn't tell them the figure doesn't exist. Both pieces of information are essential. The clearest approach isn't to hide a small icon but to state directly: "This content is a brand advertisement; the figures in the visuals are AI-generated virtual characters, not real users."

Transparency is not achieved merely because a legal team has added words somewhere on the page. Ordinary viewers must be able to understand the disclosure while it can still inform their judgment.

The Greatest Power of AI Influencers Is Not Beauty, but Unlimited Testing

Real creators have their own style and boundaries; synthetic figures can be quickly modified. Brands can test which face builds trust with a certain age group, which tone sounds like a friend, which story drives purchases best. When faces, voices, emotions, and recommendation content can all be A/B tested at scale, AI influencers aren't just endorsers—they're personalized persuasion systems.

This is especially consequential for children, teenagers, people experiencing loneliness or financial anxiety, and those seeking health information. A virtual figure can be endlessly patient, endlessly supportive, and recommend products at the most persuasive moment. Users feel companionship; platforms calculate conversion rates.

So media literacy shouldn't just teach "pictures might be fake"—it should ask further:

• Why is this character appearing before me?
• What data did the platform use to decide I'm easily persuaded by it?
• Does the recommendation reflect lived experience, or a sales script designed in advance?
• If content goes wrong, who's responsible?

Education Cannot Stop at Teaching Students to Spot Glitches

Early deepfake education often taught students to look at fingers, eyes, mouth shapes, lighting, and backgrounds. Those techniques quickly become obsolete as model quality improves. New media literacy should shift toward source and responsibility reading. Students need to learn how to examine account history, commercial disclosures, content provenance, brand relationships, and source context instead of assuming that the naked eye will always reveal an AI artifact.

The EU AI Regulatory Framework and the NIST AI Risk Management Framework both point to an important direction: content traceability and disclosure should be part of the system, not just a label slapped on at the end. Platforms, brands, and production teams can't simply say "this is new technology"; they must answer: where did the character come from? Were data from a real person used for training? Is there authorization? Are viewers' recommendations marked with interest relationships?

Civic and technology education should also let students compare: real-person endorsements, clearly labeled virtual characters, undisclosed synthetic figures, and digital avatars authorized by real people—what are the ethical and legal differences among them? The question isn't "is AI good" but "does information enable informed judgment."

Creators’ Faces and Voices Must Not Become Perpetual Leases

The AI endorsement industry also involves creator labor. Actors, models, voice actors, and influencers may be asked to license their faces, voices, and movement data so brands can generate infinite content later. If contracts don't clearly limit duration, region, product types, model training, re-licensing, posthumous use, and deletion mechanisms, a single shoot could become a perpetual personality lease.

Brands think they've bought controllable digital doubles; creators may lose the right to refuse. The most important contract terms in the future might not be how many days of shooting but "how long can the model remember me."

This will also change the creator market. Real creators may not be replaced immediately, but lower-budget commercial shoots, modeling, voice work, and editing jobs may face downward pressure on pay. If companies only pursue low costs, they'll eventually find trust is an asset too. A real creator’s value lies not only in a face, but in lived experience, judgment, reputation, and the right to refuse. Synthetic figures can copy appearance but cannot automatically generate genuine accountability.

Platforms Cannot Shift the Burden of Detection onto Users

Social platforms often reduce risk governance to user self-defense: identify suspicious content yourself, stay alert, and read the community guidelines. But when platforms themselves use recommendation algorithms to amplify content and personal data to decide persuasion methods, then ask users to protect themselves with their eyes, this allocation of responsibility is unfair.

Platforms should keep AI markers present through reposting, cropping, and recommendation flows, and provide clear information about content origin, creation time, modification records, and commercial relationships. If an account is run by a virtual character, platforms should at least make its virtual status easy to identify on the profile page and in every commercial post. Health, finance, politics, and content directed at children should face higher disclosure standards than a small, ambiguous icon.

Conclusion: A Nonexistent Influencer Still Needs a Real Accountable Party

AI endorsers aren't necessarily deception. Clearly labeled virtual characters can be very creative, lower production costs, support multilingual markets, and protect creators who don't want to show their faces.

The real danger is that responsibility disappears along with the person. A brand calls the character a tool, a platform says it merely hosts the content, and a model provider says it only generated an output; in the end, consumers alone bear the cost of being misled.

When a character asks you to believe it, like it, and buy what it recommends, you should at least have the right to know: does it exist? Who controls it behind the scenes? And who is responsible for what it says.

Sources retained from the Chinese original

• Original source: OECD｜Artificial Intelligence