«Grok, is this true?»: Patterns of AI chatbot use in online political discourse on X during the 2025 Portuguese elections

Abstract

Citizens increasingly reference AI tools in online political discussions, yet the communicative functions of these interactions remain underexplored. This study examines how Grok AI was mobilised by users on X ahead of the Portuguese legislative elections of May 2025. Using a grounded theory approach, we analysed 525 replies to posts of major political parties and leaders that mention (@) Grok AI. Most mentions were negatively framed, with only 1% showing evidence of positive tone. Beyond knowledge-building or discussions about electoral performance (eg. polling), users resort to Grok for humour and, more prominently, to moralise political actors and stereotype political identities. However, nearly half of the mentions exhibited an undetermined tone or intention, reflecting the inherent ambiguity and complexity of citizen discourse when evoking Grok as participant. Our findings point towards four distinct use models: (i) opinion validation, (ii) argumentation by proxy, (iii) constructive autonomisation and (iv) discursive and memetic framing. This study advances research on AI mediated political communication by concluding that citizens primarily appropriate Grok AI as a rhetorical resource to augment their visibility in online political debate, rather than for fact-checking purposes.