Deepfake as one of the main information threats of the 21st century

Artificial intelligence (AI) is a tool that can provide a multitude of opportunities for those who can use it effectively. However, as always with global changes in society, there is a downside—the technology is learned not only for good but also for implementing personal plans and ideas that can cause serious harm to both individuals and society as a whole. Deepfake, as a product created by AI, is currently one of the most serious information threats, as it can deceive not only an inexperienced user but also a professional in the IT field. Every day, the media landscape sees more and more high-profile examples of deepfakes that are impossible to distinguish from real audio or video recordings. The aim of this article is to present the characteristic features of deepfake as a digital product, to provide a typology of such publications based on their creation methods and objectives, as well as to describe possible markers for recognizing fakes in the online environment. The research methodology is based on a systematic-structural analysis of the media space, allowing for the typologization of deepfakes, as well as on content analysis, which made it possible to identify the goals and characteristic features of various deepfakes. The study also applies descriptive and generalization methods. The empirical basis consisted of generated content in photo, audio, and video formats that caused a public resonance and were publicly debunked either by the creators themselves, by media outlets, or by vigilant users. The chronological scope of the research is 2020–2025. A total of 69 deepfakes were analyzed. The material was collected through a comprehensive sampling method and included publications mentioned in the media (on TV channels such as “Channel One,” NTV, and “Russia 1,” the platform “Smotrim,” online publications such as “Lenta.ru” and “Gazeta.ru,” and websites of news agencies “RIA Novosti” and TASS). This approach allowed for the formation of an understanding of the themes, goals, and methods of reproducing deepfakes, as well as for identifying the most effective ways to recognize them. These markers are relevant and effective at the current stage, as they allow for the most prompt and effective distinction of generation, which is especially important in the context of the constant improvement of neural network functioning mechanisms.