Geloofwaardig nieuws in tijden van generatieve AI : een mixed-methods onderzoek naar publiekspercepties en transparantie in een AI-ondersteunde nieuwsomgeving
Year: 2026
Type: dissertation
Abstract: The starting point of this doctoral dissertation is the growing entanglement of generative artificial intelligence (genAI) with the contemporary news ecosystem. GenAI is a specific application of artificial intelligence (AI) that relies on machine learning models trained on vast amounts of data. As a result, these systems are capable of generating new, persuasive content and simulating human-like conversations. The emergence of this technology is part of a broader trend of digitalization that has fundamentally transformed journalism over the past decades. Digitalization has not only reshaped the journalistic profession, but has also changed how news is produced, distributed, and consumed. At the same time, news credibility has been under pressure for some time in a media landscape characterized by platformization, increasing polarization, the spread of disinformation, and declining institutional trust. The public emergence of genAI, following the public release of ChatGPT on November 30, 2022, has added a new layer of uncertainty to this already complex information ecosystem. As the traditional boundaries between human journalists and technology become increasingly blurred, existing interpretative frameworks for evaluating news are also being challenged. The use of genAI in journalism, therefore, raises new questions about maintaining news credibility among audiences. In both policy frameworks and scientific debates, transparency is put forward as the ethical solution to the current credibility and trust issues in journalism. This dissertation examines how the audience evaluates news when they believe that genAI has played a role in the content creation phase of news production. This perception arises from the interplay between visibility and interpretation. The audience must first notice the involvement of genAI and then assign meaning to it. This meaning-making can occur in two ways. On the one hand, news organisations can communicate the use of genAI themselves, for example through labels. On the other hand, audiences may infer genAI involvement based on credibility cues in the text or images. Previous research has mainly focused on journalistic practices, professional norms, and full automation practices. In contrast, this dissertation adopts an audience-centered approach to genAI-supported or genAI-assisted journalism. This dissertation conceptualises news credibility not as an institutional or normative ideal, but as a perception that arises among audiences, formed at the level of the individual news item. The central research question is therefore: How does the perceived use of generative artificial intelligence influence news credibility, and what role does transparency play in this relationship? Chapter 1 situates the doctoral dissertation within broader technological developments in the contemporary news ecosystem and positions it at the intersection of the AI turn and audience turn within journalism studies. This chapter discusses the broader context in which digitalization, automation, and platformization further reinforce existing tensions around journalistic authority and news credibility. In addition, the scholarly and societal relevance of an audience-centered focus is explained, and an overview is given of the research objectives and the structure of this dissertation. Chapter 2 develops the conceptual and analytical framework through which news credibility is examined in this dissertation. The framework builds on established credibility theories from the social and communication sciences (cf. Hovland & Weiss (1951) and Gaziano & McGrath (1986)), but adapts and refines them for genAI-supported and platform-driven news environments. Within this framework, news credibility is conceptualized as a judgement that is studied through the interplay of credibility cues operating at three analytical levels: the source of the news, the platform on which it is presented, and the message, including its textual and visual elements. By studying these levels within a single analytical framework, this dissertation provides an integrated lens on how audience attributes credibility in a genAI-supported news environment. In addition, the conceptualisation of transparency occupies a central place within this framework. In both policy frameworks and normative literature, transparency is conceptualised as a straightforward means of enhancing news credibility and strengthening audience trust. In this dissertation, transparency is not assumed to be a self-evident ethical principle that automatically enhances credibility or signals responsible genAI use by journalists. Instead, it is approached as a perceptual construct, in which transparency is a condition that makes the use of genAI explicitly visible to the audience. The degree of visibility may influence whether news is perceived as credible. This conceptualisation and analytical focus guide the empirical studies in this dissertation. Chapter 3 outlines the methodological choices underlying this doctoral dissertation. This dissertation adopts a post-positivist paradigm and employs a mixed-methods research design. By using a survey study, experiments, focus groups, and eye-tracking within a single doctoral dissertation, it provides a comprehensive analysis of perceptions, interpretations, and effects associated with the use of genAI and transparency. The main rationale for employing a diverse set of research methods throughout this dissertation is to examine the layered and complex nature of news credibility and genAI from multiple perspectives. Each method provides unique insights, and together they contribute to answering the research question. Chapter 4 brings together the four empirical studies of this dissertation. Each study is grounded in its own perspective and research context. These studies are embedded in the research objectives and analytical framework outlined in Chapter 2. Together, they contribute to the three overarching research objectives: (I) mapping the use of genAI applications in news reporting, (II) identifying the characteristics of news credibility, and (III) investigating the role and effects of transparency on news credibility perceptions. The first research objective focuses on mapping the use of genAI applications in news reporting, as well as the attitudes and ethical concerns associated with them. This provides insight into the current context in which audiences navigate. Study 1 addresses this through a survey of professional journalists in the Low Countries. The results show that genAI is being used in news production, primarily for text-supporting tasks such as translation, transcription, and proofreading. The findings suggest that journalists do not regard genAI in news production as disruptive but rather as a transformative force that supports and strengthens existing journalistic processes without replacing them. Journalists define their professional boundaries by determining for which tasks they do and do not deploy genAI. Regional differences (Flanders, Wallonia, and the Netherlands) remain limited, whereas individual differences are pronounced. Differences in AI literacy, in particular, give rise to distinct user and attitude patterns. Journalists themselves emphasise the importance of human control, ethical frameworks, and transparency in the use of genAI, while pointing to the absence of clear and uniform guidelines. The second research objective focuses on the characteristics of news credibility on social media. Study 2 uses an experiment to investigate how explicit source attribution, image manipulation, and platform-level transparency shape audience perceptions of news credibility on Instagram. The findings show that credibility perceptions on social media are strongly shaped by credibility cues. Source presence remains an important signal, even in visually driven environments. In contrast to platform initiatives aimed at safeguarding the authenticity of source information, platform-specific labels function as transparency mechanisms with only limited influence on credibility judgments of news. Images, by contrast, play an important role in how news is perceived. The results suggest that audiences primarily base their credibility assessments of news on social media on the extent to which images plausibly align with the news story, and whether a recognisable source is attributed, rather than on concerns about potential image manipulation or editing. The third research objective focuses on transparency about genAI use in news production. Study 3 uses focus groups to examine how audiences make sense of the role of transparency regarding genAI in journalism. Although transparency is normatively emphasised in both the scholarly literature and policy frameworks as an ethical principle, in practice, the audience gives it a task- and context-dependent interpretation, tailored to its specific information needs. The focus groups reveal that transparency is considered particularly desirable for socially sensitive topics and for core journalistic tasks such as fact-checking and multimedia production, which require nuance and human involvement. At the same time, explicit communication about genAI use can give rise to doubt or feelings of being misled. This appears to be particularly the case when the human role is insufficiently visible, and emphasis is placed too much on technological involvement rather than on the human contribution. This tension is conceptualised in this dissertation as the transparency paradox. Building on these insights, Study 4 uses an experimental eye-tracking study to examine the extent to which textual AI labels, as a transparency mechanism, draw the audience’s attention. It also investigates how this attention influences credibility perceptions. The results indicate that such labels generally receive only limited visual attention and are insufficiently processed cognitively or remembered.
Author: D'haeseleer, Stephanie
Language: English
Subfield: Communication
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7213487698
