Answers are grounded in this paper and cite their sources.
• Politicians’ and LLM-generated texts rely on apparently similar rhetorical strategies • Unlike politicians, LLMs often employ change-of-state verbs in fixed expressions • In LLM texts, presuppositions serve a narrower range of discourse functions. This study examines GPT-4’s capability to replicate linguistic strategies used in political discourse, focusing on its potential for manipulative language generation. As large language models (LLMs) become increasingly popular for text generation, concerns have grown regarding their role in spreading fake news and propaganda. This research compares French and Italian political speeches with those generated by GPT, with an emphasis on presuppositions – a rhetorical device that may subtly influence audiences by packaging some content as already known at the moment of utterance. Through a corpus-based pragmatic analysis, this study assesses how well GPT can mimic this persuasive strategy. Our findings show that, despite some apparent similarities, a closer look reveals important differences in their distribution and function compared to politicians. For example, GPT-generated texts often rely on change-of-state verbs used in fixed phrases, whereas politicians use presupposition triggers in more varied ways. Such differences, however, are challenging to detect with a ‘naked eye,’ and this represents a potential risk of LLMs in political and public discourse.
No takes yet. Share an insight, caveat, or question.
Garassino et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: