What is Real? A Systematic Review of GenAI-Enabled Artifact Creation and Perception
Luoying LIN, Sonne CHEN, Yaxuan Mao, Shengdong ZHAO
* Under Reviewing of ACM Computing Surveys
Generative AI (GenAI) amplified the age-old human tendencies to imitate, fake, and deceive into participatory, socially embedded practices. As synthetic content becomes more indistinguishable from reality, it challenges not only detection mechanisms but for understanding the human-centered dynamics—such as behaviors, intentions, and interpretations—that shape its production and reception. In response, we introduce ArtiFact — a neutral, interdisciplinary term capturing how individuals leverage GenAI to generate realistic synthetic artifacts that reshape perceptions of reality across contexts. Through a systematic review of 101 empirical studies, we propose a three-stage lifecycle of ArtiFact: Conception (e.g., motivations range from constructive to exploitative, with manipulation occurring through message, identity, and emotional vectors), Creation (e.g., shaped by psychological traits, user intent, and perceived credibility informed by source and message cues), and Reception (e.g., governed by individual heuristics and contextual factors such as emotional reactions). We outline five critical directions: (1) reframing “real vs. fake” through functional, intentional, and societal dimensions; (2) developing multidimensional labeling aligned with human sensemaking; (3) supporting upstream, intent-aware interventions; (4) embracing lifecycle interdependencies; and (5) cultivating socio-technical literacy. Together, our work provides a conceptual and empirical foundation for navigating GenAI-mediated realities and guiding adaptive, responsible engagement beyond technical detection paradigms.
LIN Luoying 