Authors: Hubert József (BCE), Bauer András (BCE)
Abstract:
Generative artificial intelligence is being increasingly used in political communication, raising new questions about both the production of political messages and their reception by social media audiences.
Our exploratory study examines whether public Facebook posts published by two central figures in the 2026 Hungarian election, Viktor Orbán and Péter Magyar, differ in their degree of “AI-likeness.” We also explore whether posts with different levels of AI-likeness are associated with different levels of audience engagement.
Facebook posts published between January and July 2026 were collected from the public pages of Viktor Orbán and Péter Magyar, covering approximately three months before and after the election. The corpus consists of 1332 posts: 609 from Orbán’s page and 723 from Magyar’s. The textual content of each post was evaluated using Google Gemini via its API. Gemini was asked to assign an AI-likeness score ranging from 0, indicating no apparent signs of AI-generated content, to 100, indicating highly AI-like content. Audience engagement was measured using the average number of likes on the posts.
Average AI-likeness scores were relatively low for both pages but higher overall for Magyar (18.2) than for Orbán (12.2). For Orbán, average engagement decreased as AI-likelihood increased. Magyar’s page showed a notably different pattern: posts in the highest AI-likelihood category received the highest average number of likes.
These preliminary findings suggest, first, that the communication strategies related to generative AI use differ. Second, that such characteristics may be associated with very different audience responses. The study has several important limitations. Most notably the small number of posts classified as highly AI-like (7 and 34 respectively), and the uncertain validity of automated AI detection. The observed relationship between AI-likeness and engagement is descriptive and does not establish causality. The results should therefore be regarded as exploratory and suggestive, providing a basis for further validation and more detailed analysis.