Generative Artificial Intelligence in Crisis Journalism: A Comparative Analysis of ChatGPT and Gemini News Texts

The growing use of generative artificial intelligence systems has raised important questions in digital journalism, particularly regarding how journalistic norms are reproduced by such systems in news production processes. This issue becomes especially significant in crisis and disaster reporting, where uncertainty is high and journalistic principles such as verification, objectivity, and ethical responsibility are critically important. This study aims to comparatively examine news texts generated by two large language models (ChatGPT and Gemini) based on the same disaster scenario. The research employs a single-case comparative qualitative content analysis. Within the scope of the analysis, the news texts produced by both models were coded and compared according to language and tone, adherence to journalistic norms, structural organization, and prompt fidelity. The findings indicate that both models generally adopt a neutral and cautious tone in news reporting. However, notable differences were observed in their textual organization. ChatGPT tends to produce a more linear structure resembling traditional agency journalism, whereas Gemini generates a modular news text organized through subheadings, which is more compatible with online journalism formats. In addition, the Gemini output was observed to expand the given context to a limited extent. Overall, the study suggests that generative large language models may reproduce journalistic norms in crisis and disaster reporting while displaying model-specific patterns in news construction. The findings also indicate that collaboration between human journalists and algorithmic systems is likely to become increasingly significant in the future of AI-assisted journalism. In this respect, the study provides a qualitative contribution to the literature on AI-supported news production.