How Journalistic Roles Shape the Adoption of Generative AI

By Lynge Asbjørn Møller, Hannes Cools & Morten Skovsgaard

The emergence of generative artificial intelligence (AI) marks a significant shift in how AI interacts with journalism. Earlier AI applications in news organisations were tailored for specific, often hidden tasks such as data analysis (Stray Citation2021), template-based news generation (Carlson Citation2015), or personalised news distribution (Gulla et al. Citation2021). Despite their impact on journalism, these AI systems often remained opaque to much of the journalistic community due to their complexity and behind-the-scenes nature (Jones, Jones, and Luger Citation2022). In contrast, generative AI tools are accessible and easy to use. Generative AI refers to machine learning systems that produce outputs mimicking human creativity by learning from extensive training data. The most common of these tools, such as ChatGPT, Gemini, DALL·E, and MidJourney, allow users to generate text or images based on simple prompts. This accessibility and versatility distinguish generative AI from earlier systems, bringing AI directly into journalists’ workflows. Journalists are already employing generative AI across various stages of the reporting process (Diakopoulos et al. Citation2024), and many news organisations have begun systematically integrating the technology into their daily news operations (Beckett and Yaseen Citation2023).

Generative AI’s potential to transform journalistic practice is likely to evoke a wide range of perspectives and experiences among journalists. With generative AI seemingly promising to streamline news production in ways previously unimaginable, some journalists may view the technology as an opportunity to enhance their work (Moran and Shaikh Citation2022). However, as generative AI begins to undertake tasks traditionally performed by journalists, others may worry not only about job displacement (Lindén Citation2018) but also about losing the sense of purpose and satisfaction that comes with their work (Olsen Citation2023). Moreover, generative AI introduces ethical challenges that further complicate its adoption, including concerns about its lack of accuracy (Hicks, Humphries, and Slater Citation2024) and potential for bias (Singh and Ramakrishnan Citation2023). This suggests that generative AI is not a one-size-fits-all solution for journalism. Instead, its adoption is likely to be shaped by the diversity of journalists’ professional identities and experiences. However, academic literature has yet to explore the influence of these factors on how journalists respond to and use generative AI. Without a deeper understanding of these differences, there is a risk of oversimplifying the responses to AI, creating a binary narrative of threat versus opportunity that ignores the profession’s inherent diversity.

Our study addresses this gap by exploring the influence of journalists’ varying role conceptions on their attitudes towards and usage patterns of generative AI. The study of professional role conceptions in journalism is a longstanding body of research that examines the different ways in which journalists perceive their normative roles in society (Hanitzsch et al. Citation2019; Mellado Citation2015). These role conceptions shape journalists’ work processes, ethical considerations, and overall approach to journalism (Donsbach Citation2008). Consequently, journalists with differing role conceptions will face distinct ethical and operational challenges when integrating AI into their work. By examining how these professional identities shape the adoption of generative AI, this research offers a more nuanced understanding of the complex dynamics at play within the profession. Such an understanding is important not only as a contribution to academic literature but also as a foundation for navigating the ongoing integration of AI in newsrooms. Failure to recognise the diverse perspectives of journalists could lead to ineffective implementation strategies and unaddressed ethical dilemmas, with broader implications for journalism’s role in society.

Specifically, we explore the relationship between generative AI and journalistic roles by drawing on a survey of Danish journalists that measures their role conceptions, self-reported uses of generative AI, and their views on benefits and concerns related to the adoption of generative AI. This approach enables us to identify and compare patterns of AI use and attitudes across four different role conceptions that emerged from the survey responses: the mobilisation role, the watchdog role, the detached observer role, and the entertainment role. This analysis finds distinct differences between the roles that largely align with traditional understandings of the respective role conceptions, suggesting that journalistic roles influence how journalists interact with generative AI. The survey is conducted in Denmark, a small but diverse media market characterised by full digitisation (Sjøvaag et al. Citation2019) and an institutionalised professionalism rooted in strong public interest traditions (Syvertsen et al. Citation2014). Danish news organisations are already investing heavily in generative AI tools that are currently being implemented in newsrooms (Newman et al. Citation2024). This provides an interesting setting to explore the early adoption of generative AI among journalists who share a commitment to high standards of journalistic practice, offering insights into variations within the profession that extend beyond the national context of the study.

Generative AI and News Automation

The public release of the text generation tool ChatGPT in November 2022 drew significant attention to the capabilities of generative AI as a highly accessible AI technology. Generative AI systems typically feature conversational interfaces that allow users to ask questions and receive responses in ways that mimic human interaction. Unlike the often hidden AI technologies in journalism, generative AI can therefore be seamlessly integrated into regular journalistic workflows, with recent research revealing use cases at every stage of the reporting process, from the gathering to the production, verification, and distribution of news (Beckett and Yaseen Citation2023; Diakopoulos et al. Citation2024). To understand the potential uses of and attitudes towards generative AI in journalism, this section links its applications to literature on journalistic responses to previous implementations of algorithms and AI in newsrooms.

One of generative AI’s core strengths lies in its ability to produce complex narrative text at scale, significantly expanding the potential for automated news writing compared to earlier template-based approaches (Nishal and Diakopoulos Citation2023). Consequently, generative AI may amplify concerns among journalists about job displacement. Studies of newsroom innovation show a trend towards increasing efficiency under financial pressure, with automation frequently adopted to reduce costs (Dickinson, Matthews, and Saltzis Citation2013). As a result, many journalists perceive news automation as a threat to their job security and satisfaction (van Dalen Citation2012). Research highlights an underlying fear among journalists that automation technologies could either replace their roles entirely (Carlson Citation2015; Jones, Jones, and Luger Citation2022; Kim and Kim Citation2018; Moran and Shaikh Citation2022) or undermine the meaningfulness of their work by reducing task and skill variation (Olsen Citation2023).

Rather than being seen solely as a threat to jobs, AI and algorithms are sometimes portrayed as opportunities to improve journalism through the automation of routine tasks (Jones, Jones, and Luger Citation2022). Generative AI effectively handles repetitive tasks such as transcription, translation, and summarisation (Beckett and Yaseen Citation2023), thereby freeing journalists to concentrate on the traditional but time-consuming tasks that define their profession (Møller, Skovsgaard, and de Vreese Citation2024). Previous studies suggest that journalists generally welcome automation technologies that reduce this kind of routine work, enabling greater emphasis on in-depth reporting and investigative journalism (Schapals and Porlezza Citation2020; van Dalen Citation2012; Wu, Tandoc, and Salmon Citation2019). In line with these studies, Moran and Shaikh (Citation2022, 1765) have found AI’s role in journalism to be pitched as a “technological liberator of journalists which can untether them from clutches of menial tasks that deprive them from producing quality work”.

Generative AI’s capacity to process, interpret, classify, and identify patterns in large and complex datasets also demonstrates its potential as an augmentation tool that enhances rather than replaces journalistic work (Nishal and Diakopoulos Citation2023). Such an augmentation approach has been found to deflate automation-related fears among journalists by ensuring that the technologies remain under human control (Bucher Citation2017; Milosavljević and Vobič Citation2019; Rydenfelt Citation2022). Moreover, it mitigates ethical concerns prevalent in the literature on automated news writing, such as maintaining standards of verification and balance (Thurman, Dörr, and Kunert Citation2017). However, generative AI introduces significant new ethical challenges. For instance, generative AI technologies are often trained on outdated data (Alkaissi and McFarlane Citation2023), they are known to consistently produce falsehoods (Hicks, Humphries, and Slater Citation2024), and they reproduce biases in their training data (Singh and Ramakrishnan Citation2023). Recent research highlights growing awareness among journalists of these ethical challenges (Diakopoulos et al. Citation2024).

This literature underscores the numerous opportunities and challenges that generative AI presents for journalists. However, a significant gap remains in understanding how generative AI might be viewed and utilised differently across various segments of the journalism profession. Current literature tends to approach its implications in broad terms, overlooking the nuanced ways in which individual journalists’ values and beliefs shape their perspectives and practices. Recognising these differences is essential for comprehending the broader implications of AI integration in newsrooms. To address this gap, our study explores how journalists view and use generative AI depending on their professional role conceptions.

Generative AI and Journalistic Roles

The ways in which journalists think about their societal roles is an important area of research in journalism studies. Central to this understanding is the concept of journalistic roles, which describe how journalists define their relationship with society and their professional responsibilities within it. Role conceptions are significant because they shape journalists’ professional behaviours and practices (Donsbach Citation2008), bridging journalism as an institution with the individual choices and actions of its practitioners (Mellado and van Dalen Citation2014). Therefore, the concept is often used to study variations in journalists’ professional identities across different contexts. Research into journalistic roles often relies on surveys or interviews, asking journalists to articulate the normative roles they believe they should fulfil within society (Hanitzsch et al. Citation2019). Building on Cohen’s (Citation1963) distinction between “neutral” and “participant” journalistic roles, this approach has enabled scholars worldwide to identify and categorise distinct role conceptions that journalists consider important for their work.

Some role conceptions are rooted in traditional normative notions of how journalism should serve the public interest. For instance, the “neutral” (Johnstone, Slawski, and Bowman Citation1972) and “disseminator” roles (Weaver and Wilhoit Citation1996) depict journalists as detached observers, dedicated to reporting facts objectively and impartially. Other normatively rooted roles adopt a more antagonistic stance toward power, such as the “watchdog” (Johnstone, Slawski, and Bowman Citation1972; Weaver and Wilhoit Citation1996) and “monitorial” roles (Hanitzsch et al. Citation2019). These roles emphasise holding authority accountable and safeguarding the public interest, often through investigative and critical reporting that uncovers hidden truths. In contrast, other roles are more audience focused. For instance, the “infotainment” (Mellado Citation2015) or “accommodative” roles (Hanitzsch et al. Citation2019) frame the audience as spectators to be entertained, addressing their needs through appealing and relatable content. Meanwhile, the “populist mobiliser” (Weaver and Wilhoit Citation1996) or “interventionist” roles (Hanitzsch et al. Citation2019) seek to mobilise the audience toward political, social, or cultural change. Journalists adhering to these roles aim to influence public opinion, advocate for change, and facilitate development.

Given their impact on behaviour, role conceptions also influence how journalists adopt emerging technologies. Previous research demonstrates that journalistic roles shape online journalist–audience relationships (Holton, Lewis, and Coddington Citation2016) and significantly affect how journalists respond to audience metrics (Belair-Gagnon, Zamith, and Holton Citation2020). Similarly, role conceptions may explain how readily journalists adopt generative AI. For instance, journalists aligned with watchdog or detached observer roles might resist generative AI due to concerns about editorial independence and the accuracy of AI-generated content. However, journalistic role conceptions are not always reflected in practice (Mellado and van Dalen Citation2014). External factors such as organisational routines, technological pressures, and precarious working conditions often constrain journalists’ ability to act in line with their values and beliefs (Tandoc, Hellmueller, and Vos Citation2013). In an era of increasing precarity in journalism, resisting new technology based on normative ideals for instance poses a greater risk of obsolescence or job displacement (Örnebring Citation2018). Consequently, journalists adhering to more audience-oriented roles may feel compelled to adopt generative AI to meet increasing content production demands.

These relationships between journalistic role conceptions and the adoption of generative AI remain largely unexplored in current literature. This article utilises journalistic role conceptions as independent variables to examine and explain variations in both journalists’ attitudes towards and usage patterns of generative AI. Specifically, we ask the following research questions:

• RQ1: How do journalistic roles influence journalists’ attitudes towards generative AI?

• RQ2: How do journalistic roles influence journalists’ use of generative AI?

Methods

To explore these research questions, this study employs a survey of Danish journalists conducted from March to June 2024. The Danish media system shares characteristics with those of the other Nordic countries, such as inclusive and diverse press markets, strong public service traditions, high journalistic professionalism, and full digitalisation (Brüggemann et al. Citation2014; Sjøvaag et al. Citation2019; Syvertsen et al. Citation2014). Moreover, the news industry in Denmark is under increasing financial pressure, and many news organisations are investing heavily in the implementation of generative AI in journalistic practices (Newman et al. Citation2024). These characteristics make Denmark a well-suited case for studying the influence of journalistic roles on the adoption of generative AI because tensions with professional ideals are more likely to emerge in a highly professionalised media system.

Surveying journalists is generally challenging due to journalism’s ill-defined boundaries complicating the definition of a target population, limited access making it difficult to reach sufficient members of that population, and time pressures inherent in journalistic work reducing their availability to participate (Molyneux and Zamith Citation2022). To mitigate these challenges, we carried out the sampling and distribution in collaboration with several major Danish media organisations, including regional media organisations Jysk Fynske Medier and Nordjyske Mediehus; national commercial television station TV2 Denmark; and national newspapers Jyllands-Posten, Ekstra Bladet, B.T., Weekendavisen, and Kristeligt Dagblad. Managers and editors within these organisations distributed the survey to their journalists. This approach allowed us to gain access to a large number of journalists and increased the visibility of our survey, ensuring that we captured the adoption of generative AI among a wide range of professional journalists in Denmark.

The survey was distributed to approximately 1,300 journalists, based on estimates provided by the managers overseeing its distribution. However, the exact number of journalists invited remains uncertain due to restricted access to the internal mailing lists of the news organisations. Of those contacted, 299 journalists responded, with 211 fully completing the survey and 88 partially completing it, resulting in a response rate of 26.3% according to AAPOR’s Response Rate 2 (AAPOR Citation2023). The sample consists of 60% men and 40% women, with most respondents being reporters (55%), followed by editors (14%) and interns (7%). Their experience in the news industry ranges from fewer than 3 years (14%) to more than 20 years (35%), with 26% having 3–10 years of experience and 25% reporting 11–20 years. While this does not constitute a representative sample of Danish journalists, it serves as a non-probability sample designed to explore variations in the adoption of generative AI. This approach is justified in cases where obtaining a representative sample is unfeasible, particularly given the substantial variation observed within our final sample on the key variables of interest in the study.

Results

The section presents the findings of the survey, focusing first on the journalists’ attitudes towards the adoption of generative AI in journalism and secondly on their uses of generative AI in their work. In each subsection, we begin with an overview of descriptive statistics before exploring the influence of journalistic role conceptions. Coefficient tables for all regression models can be found in the appendices.

For the mobilisation role, significant relationships are observed only with the importance of accountability and guidelines. Similarly, the entertainment role only shows a borderline significant positive relationship with the importance of generative AI training and a borderline significant negative relationship with the importance of transparency. These findings suggest that journalists adhering to these roles may place less emphasis on ethical considerations in implementing generative AI, potentially prioritising audience engagement and content creation over issues like accountability or transparency.

Use of Generative AI

Turning to the second research question, we examine how journalistic role conceptions shape the use of generative AI. Descriptive statistics show that over half of respondents use generative AI at least once per month in their capacity as journalists, with a quarter never using it and just over 10% using it daily. The most common applications involve brainstorming potential headlines and story ideas, while generating or editing illustrations is the least common use.

The regression analysis indicates that journalistic roles influence the use of generative AI with particularly contrasts between the detached observer and entertainment roles. These contrasting patterns likely reflect varying professional norms and values associated with each role. To illustrate these differences,