Characteristics of AI-Based Disinformation: A Comparative Content Analysis of AI and Non-AI Hoaxes Based on TurnBackHoax Mafindo Data 2025 Muhamad Handar (1)
1) Communication Studies, Indonesia Open University, East Jakarta. DKI Jakarta Province. 13230.
*muhamadhandar01[at]gmail.com
Abstract
The rapid advancement of generative artificial intelligence (AI) has opened a new chapter in disinformation production that is increasingly difficult to detect with the naked eye. This study aims to map and compare the characteristics of hoaxes produced using AI technology with conventional (non-AI) hoaxes circulating in Indonesia throughout 2025. The research employed a quantitative content analysis method on 1,544 clarification articles published on TurnBackHoax.id by the Mafindo. Each case was categorized based on theme, type of mis/disinformation, dissemination tools, channels, narrative type, evidence of claims, impersonated parties, and geographic context. The results show that of the total cases, 215 cases (13,9%) were identified as involving AI technology, while 1,323 cases (85.7%) remained non-AI. Political themed hoaxes dominated overall findings (55.2%), with Facebook as the primary dissemination channel (53.8%) and text-video combination as the most dominant tool (65.1%). AI generated content tended to leverage more convincing visual manipulation to strengthen the credibility of wedge-driver narratives aimed at fostering social polarization. The Indonesian government was the most frequently impersonated entity (37.1%), yet ironically, official clarification efforts from the government accounted for less than government, fact-checking communities, and mass media in responding to the evolving threat of synthetic disinformation driven by AI adoption, in order to strengthen Indonesia^s digital information resillience.
Topic: Strategic Communication in the Digital Sphere: Cultivating Civic Trust through Data-Driven Approaches (Digital Communication- Public Communication Strategy- Civic Trust)
