Audience Discernment of AI-Generated Media Content: A Systematic Review

Despite the growing prevalence of AI-generated content in the media industry, media research has largely focused on specific aspects, such as measuring communicators' attitudes and perceptions toward the use of this emerging technology. Research examining audiences and their ability to discern AI-generated content remains limited. Accordingly, this study aims to examine audiences' ability to discern AI-generated media content. The study employs a systematic review methodology, reviewing 19 studies based on clearly defined inclusion and exclusion criteria within the Participants, Concept, and Context (PCC) framework. The study identified several key findings. Most notably, audiences can distinguish AI-generated content through a range of technical and statistical indicators, including probabilistic and predictive indicators, digital watermarking, pixel prediction, and blockchain technologies. Recognizing these indicators requires audiences to possess an adequate level of technical knowledge and expertise. In addition, media and journalistic literacy, together with transparency and disclosure practices and linguistic indicators, can assist audiences in distinguishing AI-generated content from human-generated content. Regarding the methodological approaches adopted in previous research, the review found that studies on this topic primarily focus on issues of trust, credibility, and transparency in AI-generated content. It also revealed that questionnaires, whether electronic or paper-based, are the most commonly used data collection instrument. Furthermore, individuals constituted the most frequently studied sample, and the review highlighted the diversity of theoretical frameworks employed across the included studies.