Authors

Discussion

The total virtualization of the communicative space has led to a decrease in users' critical perception of information. The dissemination of emotional and sensitive content is not merely aimed at eliciting empathy and support from recipients. Instead, its primary purpose is to promote specific narratives and direct the audience’s attention toward desired reactions. In such cases, traditional fact-checking methods do not always prove effective. As a result, even well-informed social media users often struggle to identify false, misleading, or artificially generated content. However, at the linguistic level, such content exhibits distinct features that can help detect manipulation. 

The object of this study is to identify linguistic markers of manipulation and the use of artificial intelligence in social media content. 

The subject of the research includes social media posts containing hidden signs of AI-generated content and manipulative influence. 

The material for analysis consists of posts by English- and Ukrainian-speaking social media users published on platforms such as X, Facebook, and Telegram. The research focuses on posts related to political topics or those associated with the war in Ukraine during the period 2022–2025. Particular attention is given to multimodal posts that combine text and images, as multimodality influences audience perception and enhances emotional responses. 

The main research focus is the development of an algorithm for analyzing social media posts that will allow the detection of signs of manipulation and artificial intelligence. 

The study places particular emphasis on linguistic means at different levels: 

- Evaluative vocabulary, predominantly involving hyperbole: excellent, terrible, best, worst, etc. 

- Vocabulary from the thematic fields of "health" and "disease": ill, surgery, blood, medicine, etc. 

- Vocabulary related to "war": weapons, injury, missile, air raid, captivity, etc. 

- Vocabulary associated with "death": to die, fatality, to kill, etc. 

- Colloquial and informal language: slang, jargon, argot. 

- Euphemisms: war – conflict, aggression – liberation, etc. 

- Negative pronouns and adverbs: nobody, never, nowhere, etc. 

- Direct word order in syntactic structures. 

- Repetitive themes and narrative templates in posts. 

- Standardized text structures and AI-generated images. 

The introductory part of manipulative posts often contains emotionally charged syntactic constructions, such as: "Today is my birthday, but no one congratulated me." or "This post will not get as many likes as a picture of a naked girl." These statements often feature antithesis, creating an emotional contrast between normative and non-normative social behavior. The text is typically accompanied by emotionally evocative photographs or AI-generated images. 

The identification of manipulation and artificial intelligence in social media posts requires considering not just a single linguistic or discursive feature but an entire set of such markers. The key criterion is linguopragmatic – the speaker’s intention. Therefore, it is essential to establish the communicative intent of the content creator and those who disseminate it. If the primary intention is to provoke an emotional reaction without encouraging critical perception of the text, the post is predominantly manipulative. 

Additionally, the study analyzes the conditions under which such content spreads. It is crucial to consider the overall content of the original social media page, the authenticity of the author, possible fact-checking efforts, and user comments.