Algorithms and Authors: How Generative AI is Transforming News Production

The challenges that text generating algorithms pose as they are deployed into newsrooms and journalistic contexts loom large, from the unreliability of their outputs to the societal hype that influences their integration into news businesses. A survey conducted by The Associated Press in December 2023 revealed that nearly 70% of newsroom staffers reported using generative AI for tasks such as crafting social media posts, newsletters, headlines, and even drafting articles (Mahadevan, 2024). This paper reveals that the probabilistic modeling used by text-drafting algorithms in newsrooms is not only in direct contrast to how human journalists write articles, but is subject to opaque decisions made by technology corporations with different practices and values than news media organizations. This study assesses the experiences of journalistic practitioners through semi-structured interviews with news workers and experts, working in both technical and non-technical roles, in the United States and the United Kingdom. The results highlight how generative AI is reshaping journalism by impacting three main groups: individual journalists, news organizations, and audiences. Generative AI is transforming how individual journalists do their work and how news organizations operate–from gathering and verifying information to producing and distributing content to their audiences. First, respondents agreed that LLMs are mediocre at writing full articles on their own, but are proving useful for tasks like brainstorming, summarization, and translation. Second, at the organizational level, generative AI is pushing newsrooms and tech companies to rethink how the incentives of journalism and tech companies align. Finally, when it comes to audiences, there was less agreement from respondents on exactly how the relationship is changing, but a clear consensus that audiences are attending to the possibility of AI authorship–and its potential mistakes–in news and the fact that this could undercut trust in the accuracy of journalists. These findings underscore the importance of awareness among journalists, organizations, and the public of newsroom dependency on the technology sector given the difference between how humans and journalists create news. This paper, via the aforementioned findings, shall advance existing literature on the impact of large language models on journalism and suggest critical areas for future research on the implications of LLM integration into other areas of professional practice as well.