AI-generated presenters are already being used to sell products in livestreams. A convincing face, a fluent voice, and familiar gestures reproduce the appearance of someone speaking to an audience. The commercial possibilities are easy to see. To understand the deeper issue, however, consider a simple case: a synthetic presenter recommending a skin cream.

The presenter can describe ingredients, explain instructions, or communicate the results of a genuine test. But what happens if an invented person says, “I used this, and my skin felt better”? That figure has no skin that became irritated, no discomfort that disappeared, and no morning when it looked in the mirror and noticed a change. We are being offered an experience of using the product that never happened.

A synthetic presenter can faithfully communicate an actual person’s experience when that source is clear. A fictional character can also appear in an advertisement whose fictional nature is understood. The problem emerges when a manufactured witness is presented as someone speaking from personal experience. Even if the product itself is real, the apparent relationship between the product and the person recommending it can be invented.

Advertising has always used performances, idealized bodies, staged situations, and promises. Recognizing a persuasive intention helps us interpret what we see; researchers describe this understanding as persuasion knowledge. Recognition does not make us immune to influence, but it gives us a frame. A commercial persona woven into ordinary conversations, daily routines, and seemingly spontaneous recommendations can make that frame harder to maintain. The sales pitch arrives through the familiar gestures of everyday life.

Yet the consequences may extend far beyond buying the wrong cream. From a stream of examples, we learn what bodies look like, how relationships work, how quickly success arrives, how nature behaves, and what other people supposedly want. Those examples can become the measures we carry into the world. An invented experience can influence a real purchase; an accumulation of invented experiences may influence the world we learn to expect.

My twenty-year-old son, who grew up inside the digital world, often tells me that he approaches everything he sees online with suspicion. I take that seriously. It is a personal observation, and it raises a question that reaches beyond our conversations: what happens when a habit of suspicion learned online accompanies someone into everyday life? And before a person even decides whether something is trustworthy, what have its images already taught them about how the world works?

Much of what we know comes through other people, books, images, and screens. We cannot personally experience every place, event, or scientific process we need to understand. Digital mediation is therefore part of how we encounter reality. The difficulty grows when apparently factual examples offer convincing representations of events, lives, or experiences that never occurred, while concealing the absence of the experience they appear to report.

Albert Bandura’s account of learning from media helps explain the importance of this distinction. Through observing models, people can acquire ways of thinking and acting without undergoing the same experiences themselves. They can also extract general rules from examples and apply those rules in other situations. This process, known as observational learning and modeling, involves more than copying a particular gesture. What we observe can contribute to the standards through which we interpret later experiences.

This does not make viewers passive or their responses inevitable. People question, compare, reflect, and revise. Nevertheless, knowing that an image is artificial and noticing how it shapes our expectations are different achievements. Someone may understand perfectly well that a figure exists only on a screen, yet still use its appearance, availability, or apparent success as a reference when judging an ordinary day.

Consider what those references could contain. A body without fatigue, aging, or an inconvenient angle. A successful life from which years of effort, uncertainty, and unsuccessful attempts have disappeared. A companion who is always patient, attentive, and interested. Repeated encounters with such examples may make the ordinary limits of embodied life harder to accept. What happens to our sense of a reasonable life when its most visible models do not have to live one?

The same question reaches nature. Imagine synthetic videos in which wild animals repeatedly behave like affectionate human companions, landscapes appear perpetually spectacular, or living processes unfold without their actual conditions and timescales. If viewers treat these scenes as records of what happened, what might they learn about animals, environments, or growth? A real forest includes waiting, discomfort, decay, and long stretches when nothing remarkable happens for a camera. How might it feel to someone whose expectations were formed through uninterrupted spectacle?

Fiction and imagination can teach us a great deal. A story can deepen compassion, an animation can explain a scientific process, and a simulated environment can help someone understand a real system. The question concerns the relationship between the representation and what it claims to represent. Are we being invited into a story, shown a model with acknowledged limits, or asked to believe that an event occurred? These are different relationships with the viewer, even when their images look equally convincing.

This also helps explain why someone might follow a virtual influencer. People can feel attraction, curiosity, comfort, or identification while knowing that a figure is artificial. A beautiful image can be pleasing. A familiar voice can become part of a routine. A one-sided attachment to a recurring media figure—a parasocial relationship—can carry real feeling. Awareness of artificiality does not automatically remove the emotional value of an encounter.

We may also attribute human qualities to a nonhuman figure, a tendency called anthropomorphism. A responsive expression or an apparently considerate answer can invite us to experience personality behind the presentation. In a controlled study published in 2022, Sophie Nightingale and Hany Farid found that participants had difficulty distinguishing certain synthesized faces from real ones, and rated the synthesized faces as more trustworthy on average. Those findings concern particular static images and experimental conditions. They nevertheless illustrate how an appearance can invite a judgment it has not earned through conduct.

A viewer’s feelings can be real while the person apparently receiving them is fictional. Enjoyment alone does not establish that a recommendation comes from experience, that a response reflects concern, or that the persona has obligations toward its audience. Someone can enjoy a virtual figure as entertainment and still ask how its commercial claims were produced. The difficulty begins when the pleasure of familiarity quietly becomes a reason to trust testimony.

Perfection may eventually become less important than believable imperfection. A commercial persona could be given awkward moments, small mistakes, apparent vulnerability, and stories of disappointment. These details can make it feel more relatable. A future system might adjust its manner to what particular viewers find reassuring or attractive. This is a projection rather than a description of every current application, but it points to a serious possibility: the signs through which we recognize authenticity can themselves become material for fabrication.

The same mechanism appears without a face. A large collection of phones or accounts can generate visits that appear organic. Reviews can describe purchases that never happened. Comments can create the impression of enthusiasm among independent people. Followers and likes can suggest an audience that did not gather in the way the numbers imply. Here, the missing experience becomes a missing crowd: visible traces of collective activity without the corresponding independent participation.

This matters because other people’s apparent choices help guide our own. When uncertainty makes direct evaluation difficult, we often look for evidence of what others approve, a process commonly described as social proof. In a randomized experiment published in 2013, Lev Muchnik, Sinan Aral, and Sean Taylor found that manipulated earlier ratings influenced later voting on a social news platform. That study was not about AI influencers. It showed, however, that a displayed judgment can help produce subsequent judgments rather than simply record them.

A manufactured signal can therefore do more than decorate a page. It can shape what later visitors perceive as credible, desirable, or widely accepted. If people respond to it, some subsequent activity may become genuine, even though the initial impression was artificial. This complicates the idea that popularity is a neutral reflection of independent preference. Under some conditions, apparent popularity participates in creating the preference it seems merely to report.

For a person learning about the world through these environments, the lesson may reach beyond one product. Repeated apparent agreement can suggest what counts as success, attractiveness, intelligence, or normal behavior. The concern becomes especially serious when many supposedly independent voices are generated or coordinated by the same source. A viewer appears to be hearing from a social world, while encountering repeated versions of a much narrower set of interests.

One possible response is increasing skepticism. Another is exhaustion: if every image, voice, and recommendation requires investigation, a person may eventually stop trying to judge carefully. They might dismiss unfamiliar evidence, accept only what comes from a favored group, or trust whatever feels most emotionally plausible. These are possible responses, not a single future awaiting everyone. Still, widespread fabrication could make the work of distinguishing more demanding precisely when people have less patience for it.

Research on synthetic political video provides a limited but relevant warning. In a 2020 study, Cristian Vaccari and Andrew Chadwick found that a deepfake could produce uncertainty more readily than outright deception, with that uncertainty connected to reduced trust in news on social media. This does not demonstrate that distrust then spreads into friendships, family life, or encounters with nature. That wider movement remains a question requiring its own evidence. The study does show why counting only how many people believed a fabrication can miss part of its effect.

There is also a consequence for genuine evidence. Bobby Chesney and Danielle Citron describe the “liar’s dividend”: awareness that recordings can be fabricated can make it easier to dismiss authentic material as fake. A real event acquires an additional burden of proof because convincing false events are possible. Under those conditions, better fabrication can weaken the standing of evidence that has nothing artificial about it.

Could a similar habit reach face-to-face life? Could someone become quicker to interpret kindness as strategy, vulnerability as performance, or testimony as a prepared script? We should resist presenting that movement as an established consequence of AI. But we should also resist treating the screen as a sealed environment whose habits end when it goes dark. People carry learned expectations into encounters, and those expectations can affect what they notice, how they interpret it, and how they respond.

The risk may involve both misplaced belief and misplaced disbelief. A person can absorb unrealistic standards from artificial examples while becoming suspicious of authentic experiences that fail to match them. They might expect effortless achievement yet doubt an honest account of struggle. They might expect constant availability yet interpret an ordinary boundary as indifference. This combination could leave reality facing two demands at once: resemble the convincing image, and prove that it deserves to be believed.

There may also be a feedback process. People who see particular performances rewarded online may reproduce them in everyday life. Those real performances can then be recorded and circulated as further examples of how people naturally behave. An initially constructed pattern acquires a lived expression, which appears to confirm the pattern. This is an interpretation of a possible mechanism, not proof that society is moving uniformly in that direction. It nevertheless asks us to consider how representations and conduct can shape one another.

Jean Baudrillard’s idea of hyperreality offers a philosophical lens for this problem: representations can become references through which the world they represent is judged. Applied here, the question is whether a synthetic image of life begins to set the terms on which actual life appears adequate. A relationship, a body, an animal, or a landscape may then seem disappointing because it fails to behave like its more convincing representation.

At that point, the issue concerns our understanding of what is real, natural, possible, and normal. A plausible image can make something seem possible without showing that it is. Repetition can make something seem common without establishing how often it occurs. A confident account can sound like testimony without having a witness behind it. These distinctions matter because we make real decisions with the expectations we derive from what we encounter.

How, then, can those expectations remain open to correction? Direct experience helps, but it is also limited and fallible. We need other people, careful reporting, independent investigation, scientific methods, and relationships whose reliability can be assessed over time. The useful question is what supports a particular claim: an identifiable experience, a reproducible test, independent observations, or a presentation whose underlying basis remains unclear. A realistic face cannot answer that question for us.

Clear disclosure can help establish what kind of figure we are encountering. It also leaves further work to do. Knowing that a presenter is synthetic tells us something about its construction; it does not tell us whether its claims are accurate or whether its apparent experience has a real source. Similarly, knowing that an image is fictional does not necessarily reveal every expectation we have learned through watching it.

Perhaps awareness needs to include this second level of attention. We can ask whether a scene happened, and also ask what it taught us to expect. What kind of body now seems ordinary? What pace of success seems reasonable? What behavior do we expect from animals, friends, or partners? Which frustrations have we begun to interpret as failures because the examples against which we judge them contained no comparable limits?

These questions apply to this essay as well. Research findings, a personal observation, and a future projection offer different kinds of support. AI assistance in developing and expressing an argument cannot supply a firsthand experience that the author has not had, or turn a plausible interpretation into an established result. A text that asks readers to examine appearances owes them clarity about the grounds of its own claims.

We began with a synthetic presenter recommending a product. Seen from this wider perspective, that scene may also contribute to a model of bodies, experience, credibility, and ordinary life. Its immediate commercial purpose is only part of what a viewer can learn from it.

When we learn about the world through synthetic lives, experiences, and events presented as though they had happened, what will we come to accept as real, natural, possible, and normal? And when we step away from the screen, which of those measures will we continue to live by?

**Sources**

- JD Corporate Blog. Announcement on the AI digital representative used in Richard Liu’s livestream debut, 22 April 2024. - Marian Friestad and Peter Wright. *The Persuasion Knowledge Model: How People Cope with Persuasion Attempts.* Journal of Consumer Research, 1994. - Albert Bandura. *Social Cognitive Theory of Mass Communication.* Media Psychology, 2001. - Sophie J. Nightingale and Hany Farid. *AI-synthesized faces are indistinguishable from real faces and more trustworthy.* Proceedings of the National Academy of Sciences, 2022. - Lev Muchnik, Sinan Aral, and Sean J. Taylor. *Social Influence Bias: A Randomized Experiment.* Science, 2013. - Cristian Vaccari and Andrew Chadwick. *Deepfakes and Disinformation: Exploring the Impact of Synthetic Political Video on Deception, Uncertainty, and Trust in News.* Social Media + Society, 2020. - Bobby Chesney and Danielle Citron. *Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security.* California Law Review, 2019. - Jean Baudrillard. *Simulacra and Simulation.* English edition, 1994.