Media as an element of the infrastructure of public verifiability: the transformation of journalism functions in the context of the proliferation of synthetic content

The subject of the research is the representation of media resources and mediated platform texts as sources accompanying the responses of generative artificial intelligence and their role in ensuring the public verifiability of machine knowledge. The study examines the transformation of journalistic functions as an expansion of traditional informational, interpretational, and control functions: the media texts produced by editorial offices become a source base, means of attribution, and elements of public verification for the responses of generative artificial intelligence. A 4 × 4 matrix has been developed, combining four modes of knowledge availability concerning AI memory – long-standing stable, recent documented, marginal, and new event-based – and four functional types of questions: “What?”, “Why?”, “How?”, and “Who?”. The matrix demonstrates the usage of media resources in the responses of generative systems. A comparative study of the responses of four generative systems – GigaChat, DeepSeek, ChatGPT, and Perplexity – was conducted using free user versions. The overall questionnaire included 80 questions, and the research set comprised 320 responses. Content analysis, source typology, double coding, and statistical verification of relationships using Pearson's χ² criterion were applied. In the 320 responses, 423 direct references were identified. The frequency of external source presentation significantly varied across platforms: from 30.0% of responses in GigaChat to 100% in DeepSeek. It was established that the representation of links to media resources primarily depends on the mode of knowledge rather than the functional form of the question. In the new event-based mode, references to media resources appeared in 55.0% of responses; in the recent documented mode, 51.2%; in the long-standing stable mode, 25.0%; and in the marginal mode, 8.8%. The relationship between the mode of knowledge and the presentation of links to media resources is statistically significant (χ² = 51.099; p " 0.001; Cramer's V = 0.400), while the differences between the questions “What?”, “Why?”, “How?”, and “Who?” are not significant. The function of media as an infrastructure for the actualization and public verifiability of synthetic content is justified. It has been established that in the generative environment, journalism acquires an additional infrastructural function – the production of attributed, verifiable, and accessible knowledge.