Why Deepfakes Spread: New Study Links Adoption to Usefulness and Fun

Deepfakes have long been framed as one of the defining ethical hazards of the synthetic media era, a technology whose capacity for deception has prompted legislation, detection research, and widespread public alarm. Yet people keep using these tools, and in many cases they keep using them even while acknowledging the very risks that make the technology so controversial. A new study published in Information Systems Frontiers tackles this paradox head-on, and its central finding is striking: ethical concern does not uniformly suppress the adoption of deep synthesis technology. Instead, the moral calculus that governs whether people embrace these applications changes fundamentally depending on whether the application is built to be useful or built to be entertaining.

The research, conducted by You Wu and Xing Zhang of the School of Management at Wuhan Textile University and Yingzhen Feng of Hubei Vocational College of Railway Transportation, integrates the Unified Theory of Acceptance and Use of Technology, widely known as UTAUT, with the Hunt-Vitell general theory of marketing ethics. UTAUT is one of the most influential frameworks in information systems research, explaining technology adoption through constructs such as performance expectancy, the belief that a technology will improve task performance; effort expectancy, the perceived ease of use; social influence; and facilitating conditions. But as the authors argue, conventional adoption models like UTAUT lack ethical constructs altogether, which may render them insufficient for explaining adoption in morally contested technological contexts such as deep synthesis technology, or DST. The Hunt-Vitell model, by contrast, treats ethical decision-making as a process in which individuals perceive an ethical problem, evaluate alternatives against deontological and teleological norms, and then form moral judgments that feed into intentions.

To capture both dimensions, the researchers designed a scenario-based survey that distinguished between two broad categories of DST applications. Utilitarian deep synthesis contexts were function-oriented, encompassing uses such as public interest communication and organizational management, where synthetic media serves instrumental purposes like conveying information efficiently or streamlining administrative work. Hedonic contexts, by contrast, were entertainment-oriented, including celebrity impersonation and virtual performance, where the appeal lies in novelty, fun, and sensory pleasure. Participants evaluated their adoption intentions across these scenarios, and the study then applied a multi-method analytical strategy combining comparative analysis with configurational analysis, specifically fuzzy-set qualitative comparative analysis, a technique capable of identifying multiple, equally effective pathways, or equifinal configurations, of conditions that lead to the same outcome.

The comparative results were unambiguous. Adoption intentions were significantly stronger toward utilitarian DST than toward hedonic DST. This finding alone challenges the popular narrative that deepfakes spread primarily because they are fun; on the contrary, the function-oriented applications of synthetic media appear to command considerably more willingness to engage. But the deeper insight emerged from the configurational analysis, which revealed that the two adoption contexts operate according to fundamentally distinct causal logics.

Utilitarian DST adoption follows what the authors describe as a rational-instrumental logic. Across every configuration leading to utilitarian adoption, descriptive norms served as the core condition. Descriptive norms refer to perceptions of what other people are actually doing, as opposed to injunctive norms, which capture what other people approve or expect. In other words, the single most important driver of utilitarian deep synthesis adoption is the simple observation that friends, colleagues, and online communities are already using the technology. Performance expectancy, effort expectancy, and technology self-efficacy, an individual’s confidence in their own technical skills, also emerged as frequently recurring core conditions across multiple pathways. The picture is one of pragmatic calculation: people adopt functional synthetic media because it works, because it is easy to use, because they feel competent handling it, and crucially, because everyone around them seems to be doing it.

Hedonic DST adoption tells a completely different story. Here, the governing framework is an ethico-normative logic, with ethical acceptability serving as the core condition across all configurations. When it comes to entertainment-oriented deep synthesis, whether people find the technology morally tolerable is not merely one factor among many; it is the pivotal condition on which adoption hinges. Alongside ethical acceptability, injunctive norms, hedonic performance expectancy, which measures how fun, novel, and pleasurable the experience feels, and the absence of deepfake concern appeared as core conditions across several pathways. This last condition is particularly revealing. Whereas utilitarian adoption can proceed regardless of nagging worries about deepfakes, hedonic adoption is far more sensitive to those concerns. People will readily use synthetic media for functional purposes even while harboring anxiety about manipulation and misinformation, but they will only embrace it for entertainment when those anxieties are absent or muted and when they have judged the use to be ethically acceptable.

The study’s measurement instruments underscore how these constructs were operationalized. Deepfake concern was assessed through items probing fears of falling for a deepfake, becoming a victim of harmful synthetic media, the erosion of trust in video content, the blurring line between reality and fiction, and the spread of misinformation. Ethical acceptability was measured through statements about whether DST applications are ethically tolerable and whether resisting them is necessary. Hedonic performance expectancy captured whether using the technology is fun, enjoyable, novel, and pleasurable. This granular approach allowed the researchers to disentangle the distinct roles that cognition, social context, and morality play in different adoption settings.

The theoretical implications are substantial. By demonstrating that ethical risk does not uniformly suppress adoption behavior, the study forces a revision of how researchers think about technology acceptance in ethically sensitive domains. The causal role of ethics varies systematically across application contexts: it is largely peripheral in the rational-instrumental logic of utilitarian adoption but constitutive of the ethico-normative logic governing hedonic adoption. This finding extends UTAUT into territory the original model never contemplated and suggests that any adoption model applied to morally contested technologies must explicitly incorporate ethical constructs rather than treating ethics as noise or assuming it operates identically everywhere.

The practical consequences for governance and regulation may be even more significant. The authors argue that their results support differentiated, evidence-based, context-sensitive regulatory design rather than blanket approaches. If utilitarian adoption is driven by descriptive norms and pragmatic expectations, then interventions aimed at those contexts might focus on transparency requirements, professional norms, and modeling responsible use within organizations and communities, since people are clearly following the crowd. If hedonic adoption is gated by ethical acceptability and diminished concern, then entertainment-oriented deep synthesis calls for interventions that cultivate ethical judgment and realistic awareness of deepfake risks, because in that domain, the absence of concern is precisely what enables engagement. Regulators who apply the same policy instruments to both contexts, the study suggests, are likely to miss the mark in at least one of them.

The work also resonates with a broader research conversation on deepfakes and synthetic media. Prior studies have examined the effects of political deepfakes embedded as vox populi on social media, the psychological processes underlying the reception of deepfakes as narratives, and how falling for a deepfake changes the way people see, hear, and experience media. Others have analyzed disparities in deepfake knowledge and attitudes, explored the motivations for sharing political deepfake videos through the lens of moral consciousness, and investigated what drives the ethical acceptance of deep synthesis applications using similar configurational methods. The new study builds on this foundation but breaks fresh ground by treating DST as a heterogeneous category rather than a monolith, and by showing that the heterogeneous role of ethical factors is the key to explaining why the technology spreads.

Funded by the National Natural Science Foundation of China and several regional research programs, the study arrives at a moment when synthetic media is becoming embedded in everyday digital life, from organizational communications to streaming entertainment. As the technology grows more capable and more accessible, understanding not just whether people adopt it but why, and under what moral and social conditions, becomes an urgent task for both scholars and policymakers. This research offers a compelling answer: the reasons deepfakes spread are not singular but dual, split between a rational-instrumental logic powered by social proof and pragmatic utility, and an ethico-normative logic powered by moral judgment and pleasure. Any serious effort to govern deep synthesis technology, the authors conclude, must reckon with both.