The Deepfake Doctor Problem: Why Labeling Synthetic Experts Is Necessary but Insufficient for Protecting Consumer Trust

Abstract

Recent advances in generative AI have made it possible to create highly realistic virtual experts for use in commercial persuasion (Mogaji & Jain, 2024). In advertising, synthetic faces or voices can now be designed to resemble doctors, professors, and other authoritative figures with increasing ease (Chuan et al., 2023). Among the most consequential of these are synthetic experts, defined here as AI-generated endorsers whose persuasive function depends on simulating recognizable markers of professional expertise even when no corresponding human expert stands behind the message. Unlike fictional characters or virtual influencers, synthetic experts are persuasive because they are designed to appear professionally credible enough to activate socially learned judgments about who is knowledgeable, trustworthy, and qualified to recommend a product. This development raises a problem that extends beyond ordinary concerns about misleading claims. When the appearance and performance of expertise can be algorithmically manufactured, consumers may respond not only to what is being said, but also to a carefully constructed impression of who is speaking. The growing use of AI-generated endorsers (Kim et al., 2025;Hong et al., 2025) therefore calls for closer attention to how trust, credibility, and authenticity are negotiated in digitally mediated advertising environments.A timely example of this challenge can be found in the Korea Fair Trade Commission's recent proposal (2026) to revise its advertising review guidelines to require disclosure when AI-generated virtual figures are used in recommendations and endorsements. It suggests growing regulatory attention to whether consumers can identify the status of the speaker in AImediated endorsements. The proposal is significant because it recognizes that consumers may mistake synthetic experts for real professionals and, on that basis, make decisions they would not otherwise make (Yoo et al., 2025). This article takes that regulatory development as a point of departure and argues that labeling synthetic experts is both necessary and justified as a minimum condition of source transparency (Burrus et al., 2024). At the same time, it contends that disclosure alone is insufficient for protecting consumer trust (Kim & Wang, 2023, 2024). Synthetic experts do not persuade only because their artificiality is hidden. They also persuade because they reproduce the visual (Wang et al., 2026) and symbolic (Bansal et al., 2024) cues through which expertise is socially recognized (Kim & Kong, 2023). The problem, therefore, is not merely whether consumers are told that an endorser is artificial, but whether such disclosure meaningfully disrupts the credibility effects of simulated authority. Synthetic experts should be distinguished from the fictional characters or virtual influencers currently used in contemporary advertising. Their persuasive force depends on appearing professionally credible enough to activate socially familiar judgments about expertise (Yoo et al., 2025). In everyday communication, audiences rely on paralinguistic cues (e.g., appearance, tone of voice, age, and technical vocabulary) when assessing whether a speaker seems trustworthy or competent (Scherer et al., 1973). Generative AI can now reproduce these cues with remarkable precision (Cao & Wang, 2025), allowing advertisers to construct figures that do not merely deliver a message but perform authority itself. What is at stake is not only the truthfulness of an advertising claim, but the strategic fabrication of a credible source.This source problem is especially important in endorsement advertising, where persuasion often depends as much on the perceived identity of the endorser as on the content of the recommendation itself (Munnukka et al., 2016). Consumers routinely interpret endorsements through assumptions about who is speaking, what kind of knowledge that person possesses, and whether the recommendation is grounded in genuine expertise or experience. A doctor recommending a health-related product, for example, carries a different persuasive weight than an ordinary consumer or a fictional spokesperson because professional authority serves as a shortcut for credibility (Nicholas et al., 2003). From this perspective, labeling synthetic experts is normatively necessary because it restores a minimum level of transparency about the status of the speaker and helps preserve the conditions for more informed consumer judgment. Yet the presence of a label does not guarantee that consumers will revise their judgments. Research on persuasion has long shown that audiences rely on heuristic cues when evaluating messages (Chaiken, 1980), especially in fast-moving or visually rich environments (Joo et al., 2026), where source credibility cues may shape judgment before viewers fully reflect on the persuasive source. In such settings, disclosure may be formally present yet cognitively weak because it is too brief, too small, poorly placed, or overshadowed by the synthetic expert itself. Its insufficiency may operate at several levels. Consumers may fail to notice the disclosure, notice it without understanding its meaning, understand it while still relying on professional authority cues, or receive transparency without any clearer accountability for the claim. This concern is especially important in video advertising, where text-only labels may be less salient than a synthetic expert's voice, appearance, and professional setting, suggesting the need for more prominent visual or verbal disclosures in high-risk contexts. The regulatory question, therefore, is not only whether disclosure exists, but whether it can meaningfully interrupt the persuasive force of simulated authority.Even when viewers do notice a disclosure, the persuasive force of synthetic expertise may not disappear. This is because credibility judgments are not formed solely through conscious assessments of factual information, but also through familiar visual and symbolic cues that continue to shape interpretation after artificiality is acknowledged (Gugerty & Link, 2020). In health-related advertising, for example, a figure who looks and sounds like a doctor may still retain the aura of professional authority even when labeled as synthetic (Nicholas et al., 2003). The problem, then, is not simply that audiences may fail to see the disclosure, but that disclosure alone may be unable to neutralize the broader communicative power of simulated expertise. In such cases, transparency remains necessary, but it should not be mistaken for a complete solution to the problem of consumer trust. This limitation matters because the core issue is not exhausted by whether consumers know that an endorser is artificial. Disclosure addresses the status of the speaker, but not the social basis of the speaker's authority. A synthetic expert may be recognized as AI-generated and still remain persuasive if it successfully reproduces the cues through which expertise is ordinarily inferred. The deeper concern, then, is that generative AI allows the appearance of professional credibility to be separated from the training, affiliation, and public accountability that normally give expertise its social legitimacy. The significance of the Korea Fair Trade Commission's proposed revision (2026) extends beyond disclosure as a formal regulatory requirement. Its broader importance is that it draws attention to a deeper problem in AI-mediated advertising: synthetic experts can project authority without being subject to the forms of accountability that ordinarily make expertise trustworthy. In conventional expert communication, credibility is attached not only to the content of a message but also to a socially recognizable speaker whose authority is presumed to rest on training, affiliation, professional norms, and the possibility of public scrutiny (Little & Green, 2022). Synthetic experts disrupt that relationship by projecting competence without responsibility. What makes this development consequential is not only that no actual expert may stand behind the recommendation, but that expertise itself can now be designed as a persuasive surface effect.Generative AI allows advertisers to assemble recognizable markers of expertise without any necessary connection to an actual qualified person. This matters because expert communication has traditionally depended on a link between representation and responsibility (Robinson, 2009). A physician's recommendation carries persuasive weight not only because it sounds credible, but because it is understood to come from a real professional who can, at least in principle, be identified, evaluated, and held accountable. Synthetic experts weaken that link by transforming expertise into a set of reproducible surface cues. In this sense, they do not merely imitate experts. They alter the communicative basis on which expertise is judged.For this reason, an effective response to synthetic experts cannot rely on disclosure alone. Regulatory transparency remains necessary, but it should be complemented by safeguards that address the credibility and accountability of synthetic expertise. These may include more prominent disclosure formats in high-risk contexts, clearer identification of the advertiser or organization responsible for the claim, restrictions on synthetic personas making first-person experiential claims, and stricter scrutiny when professional authority is used in health or financial advertising. AI literacy also remains important because audiences need to understand not only whether a face, voice, or persona has been artificially generated, but also how synthetic media mobilizes visual, vocal, and symbolic cues to simulate authority and shape trust. Protecting consumer trust, therefore, requires not only labeling rules, but also broader efforts to strengthen critical judgment in AI-mediated advertising environments (Narayan et al., 2026). This article has argued that synthetic experts pose a challenge that cannot be reduced to false claims or inadequate disclosure. As an opinion article, this argument should be read as a conceptual and normative concern rather than as direct empirical evidence that all synthetic experts remain persuasive after disclosure. Disclosure remains a necessary baseline for source transparency, but it cannot fully address how synthetic experts reproduce the visual, vocal, and symbolic cues through which expertise is socially recognized. The deeper problem is the growing ability of generative AI to detach the appearance of expertise from the accountability that ordinarily makes expertise credible. As the Korea Fair Trade Commission's proposed revision suggests, disclosure is an important first step, but it must be paired with broader attention to credibility, accountability, and AI literacy.Future research should examine whether the "deepfake doctor problem" is one instance of a broader shift in AI-mediated communication (Hancock et al., 2020), in which social legitimacy is increasingly performed through synthetic personas rather than embodied human actors. In gaming environments (Zeng et al., 2026;Kim, 2025), AI-generated characters may shape how users interpret identity, authority, and authenticity. In human-AI collaboration (Sun et al., 2025), designed cues of competence and confidence may influence trust in artificial partners and decision-support systems. In the context of social bots (Fischer et al., 2023) and social media (Wang et al., 2025), further work is needed to understand how automated personas generate credibility and social influence without corresponding human accountability. Taken together, these contexts suggest that the deepfake doctor problem is not an isolated issue in advertising, but part of a broader transformation in how expertise, trust, and legitimacy are performed in AI-mediated communication.