Since ChatGPT's release in 2022, public discourse has surged around AI's promises and risks. News coverage influences public understandings of AI, yet few studies examine its framing as a threat. Portrayed like immigrants, AI appears as an "artificial immigrant" within human-technology relations, posing threats to jobs and humanity. Guided by media framing and Integrated Threat Theory, analyzing 1,005 articles from four U.S. outlets reveals five main topics. Realistic threats (material/physical) are emphasized over symbolic ones (identity-based). Multiple correspondence analysis indicates that tech-related coverage skews toward realistic framing, while labor and creativity coverage encompass both, though associations are not statistically significant.
Extending diverse strands of scholarship on technology, AI, and mass/social media, we advance a multilevel framework for understanding trust in generative AI (GAI). This framework comprises three propositions: 1) trust is compartmentalized, with trust at output, agent, system, provider, and phenomenon levels anchored by distinct primary antecedents; 2) trust can spill over across levels, with trust in GAI more transferable than trust in mass/social media; 3) whether trust remains compartmentalized or spills over across levels depends on individual and contextual factors. The framework provides a theoretical foundation for future research on GAI and for designing GAI literacy interventions, systems, and policies.
As AI becomes more widely used in journalism, it remains unclear whether AI-generated news reduces or reinforces bias compared with human-written news. Using 18,282 AI-generated and 18,210 human-written articles, this study analyze three dimensions of bias: coverage, gatekeeping, and statement bias. Results show that AI reproduces human patterns in gender and racial representation. In topic selection, AI emphasizes high-visibility events and produces more compartmentalized structures, while human-written news exhibits more hybridized coverage. AI-generated content also demonstrates higher levels of linguistic expressive bias and greater variability in sentiment polarization, whereas clickbait levels are similar. Overall, AI restructures rather than systematically reduces media bias, functioning as a mirror and selective amplifier of existing patterns.
This study investigates how mobile app features (AI transparency, rewards, and tracking) influence smart bin adoption. Grounded in nudge theory and the Stimulus-Organism-Response framework, an online experiment tested their effects on trust, perceived value, and adoption intention. Trust mediated the effect of all three features on adoption intention. Perceived value also mediated the effect of AI type and rewards, but only white-box AI showed a direct effect on intention. An interaction effect between rewards and tracking was found in the system with black-box AI. The findings offer theoretical contributions and practical guidelines for promoting sustainable technology use.
As artificial intelligence (AI) is increasingly deployed in fact-checking, questions remain about how audiences perceive AI-generated verdicts and the sources used to justify them. Drawing on motivated reasoning and research on source credibility, this study examines how fake news congruence and AI-cited source congruence jointly shape belief correction and message- and agent-level perceptions. In a U.S.-based 2 * 3 online experiment (N = 682), participants evaluated partisan misinformation followed by AI-generated verdicts citing politically congruent, incongruent, or third-party sources. Results show that while partisan congruence of fake news systematically shaped perceptions of AI verdicts and agents, source congruence operated conditionally, enhancing corrective effectiveness under co-directional configurations (citing in-group source to debunk in-group fake news). Findings highlight the role of source citations and expectation-based mechanisms in AI-generated fact-checking.
This study examines how artificial intelligence (AI) reshapes epistemic authority in contemporary journalism. Drawing on 75 in-depth interviews with journalists, editors, media owners, and AI developers in the United States, it argues that AI does not simply automate journalistic tasks or redistribute professional roles. Instead, AI produces a condition of epistemic dislocation, in which the production, validation, and control of knowledge become fragmented and misaligned across human actors, technological systems, and organizational structures. The analysis identifies three interrelated dynamics: (1) the externalization of epistemic authority despite centralized organizational decision-making, (2) the coexistence of increasing epistemic delegation to AI and the performative reassertion of professional journalistic authority, and (3) the decoupling of knowledge production from accountability through fragmented governance structures. The present study advances a new conceptual framework for understanding AI journalism and offers implications for professional authority, ethical governance, and accountability in increasingly AI-mediated news production.
As tone selection has become an option in AI chatbots, understanding its effects on persuasion and information processing is important. Focusing on flattering tones in AI fact-checking, we investigated how this style impacts tool perceptions and information engagement. Unlike sycophancy, which distorts truth, a flattering tone enhances users' self-esteem through praise and compliments. Participants completed a brief trivia quiz and were then asked to verify their answers using a pre-programmed AI factchecking tool. We found that fact-checks delivered in a flattering tone were preferred for future use yet were associated with reduced message recall and reference checking. These effects did not depend on whether their initial answers were correct.
This study examines how Artificial Intelligence (AI) is integrated into Arabic-language fact-checking ecosystems through semi-structured interviews with 12 practitioners and comparative case analysis of four organizations: Sanad, Fatabyyano, Sawab, and Misbar. Drawing on practitioner-reported routines and organizational practices, we analyze how AI supports verification under dialectal variation, geopolitical sensitivity, and uneven digital literacy. Results reveal a shift toward hybrid human-in-the-loop configurations in which AI assists triage and pattern detection while interpretive judgment and verification authority remain human-led. Organizations consistently prioritize procedural transparency, contextual accuracy, and institutional independence over speed and scale. These findings advance debates on algorithmic accountability and human oversight while contributing to scholarship on verification systems in the Global South. This study highlights the need for culturally aligned approaches to automated fact-checking in politically complex and multilingual information environments.
In their everyday work, journalists are tasked with detecting and correcting misinformation. As technology advances, these efforts have become increasingly reliant on automation. Research on journalism's "debunking turn" focuses on AI's role in fact-checking routines; content analyses uncover frames about AI within news stories. Yet scholarship rarely treats journalistic fact-checking itself as a framing practice. Embracing this view, this article reviews the historical context in which framing has been conceptually entangled with misinformation ("challenges"). Using current news items as exemplar cases, we then articulate four principles by which fact-checking implicates news framing processes ("cues" and "coverage"). Finally, we outline ways to study fact-checking as news framing, focusing on efforts to enhance journalistic literacy ("culture").
A total of 780 Saudi Olympic viewers were surveyed immediately following the 2024 Paris Summer Olympic Games to determine the degree to which Social TV use was prevalent within the population. This study situates Social TV within the Saudi Arabian sport media context, where mobile media, Olympic fandom, and national identity increasingly intersect during major international sporting events. While social presence predicted one's likelihood to participate in Social TV behaviors, bridging social capital and perceived sociability did not. Additionally, age was inversely related with Social TV behaviors, suggesting that younger viewers were more likely to incorporate interactive and second-screen practices into their Olympic viewing routines. Patriotism and one's identification with the Saudi Olympic team predicted Social TV use while internationalism, or the desire for global kinship, was an inverse predictor. Together, these findings extend Social TV research beyond Western media contexts and highlight how socially interactive Olympic viewing may be shaped by both media motivations and culturally specific identity orientations.
This study examines U.S. news audiences' trust in AI-generated journalism across three dimensions: editorial workflow stages, computational data structure (structured vs. unstructured), and geographic scope (local vs. national). Using a nationally quota-sampled survey (N = 1,506), we find that audiences trust AI-generated journalism involving structured data (e.g. weather reports, election results) more than unstructured data (e.g. opinion pieces, feature stories). Trust in local and national newsrooms predicts trust in AI journalism, while minimal variation across editorial workflow stages suggests audiences evaluate AI journalism by output type rather than production process. These findings establish baseline evidence for future research on AI use in journalism and media transparency.
Artificial intelligence (AI) is transforming how people consume information online, raising ethical concerns about misinformation, accountability, and trust. As AI-generated content proliferates, communication leaders are increasingly tasked to verify content, making fact-checking an ethical and professional responsibility. This study examines how ethical decision-making and perceived credibility shape communication managers' (N = 335) use of AI-powered verification tools. The study's findings show AI ethical anxiety, a facet of ethical decision-making, predicts fact-checking behaviors. Perceived credibility moderates the relationship between ethical anxiety and use of fact-checking websites but not AI-powered fact-checking. Findings highlight the importance of understanding the ethical drivers behind AI fact-checking practices.
AI has been increasingly deployed to combat misinformation, yet its credibility remains contested, partly due to overlooking the factor of corrective strategies. This study addresses this gap by conducting a 3 (source: human vs. AI vs. human-AI hybrid) x 2 (corrective strategy: evidence-based vs. authority-based) x 2 (topic: social vs. scientific) mixed-design experiment. Results showed that AI adopting authority-based strategies was associated with decreased credibility, while stronger machine heuristic beliefs and authoritarian personality mitigated this negative effect for AI and hybrid sources. Findings highlight the confounding role of corrective strategies in AI credibility scholarship, offering implications for automated misinformation correction.
This essay examines how humans interface with AI, misinformation, and fact-checking through the lens of truth-default theory. Readers are introduced to Truth-default Theory and its implications for human-AI interaction and fact-checking. The essay begins with descriptions of students using hallucinated references and of a recent series of experiments using AI to detect deception. Truth-default theory is summarized. According to the theory, most information is sufficiently accurate, and when it is not, fact-checking is the gold standard for assessing verisimilitude. Absent a trigger, however, people do not consider veracity, creating human vulnerability to misinformation. Implications for AI, misinformation, and algorithmic fact-checking are discussed
The proliferation of algorithmically curated and personalized news environments has raised important questions about how adolescents encounter and engage with information. This study examines the associations between algorithmic news consumption, perceived algorithmic literacy, and multiple dimensions of news-related orientations among teenagers. Drawing on a nationally representative survey of U.S. adolescents aged 14-17 (N = 437), the study adopts a multidimensional approach to news literacy, encompassing cognitive, motivational, evaluative, affective, and knowledge-based components. Findings indicate modest positive associations between algorithmic news use and several news-related orientations, including news appreciation, need for cognition, and locus of control, while relationships with critical evaluation and knowledge are more limited. Algorithmic literacy is also strongly associated with these orientations, although substantial conceptual overlap is observed. The results suggest a conditional and non-deterministic relationship between algorithmic media use and youth news engagement. Implications are discussed with attention to measurement constraints, cross-sectional design, and the need for more precise conceptual and methodological approaches in future research.
Employing an online experiment, this study explores how awareness of AI bias could affect users' intention to adopt generative AI. Results showed a three-way interaction effect between bias-based engagement, bias-awareness information cue priming, and social norms on AI trust, which could further affect users' adoption intention. That is, with high social norms of AI use, if users had not been primed that AI was biased, their trust in AI decreased after experiencing AI bias on their own. However, if they had been primed that AI was biased, their own experience with AI bias did not significantly affect trust in AI.
The increasing adoption of generative artificial intelligence (GenAI) is reshaping how individuals seek health information. However, its use in health information seeking (HIS) raises concerns about AI hallucination risks (i.e., inaccurate or fabricated outputs) and privacy risks related to personal data misuse. Despite these concerns, little is known about how such risk perceptions are formed or how they relate to users' attitudes and continued use intentions. Drawing on the social amplification of risk framework and the concept of machine heuristics, we conducted an online survey of 1,038 Chinese users who had used GenAI for HIS. Results showed that greater attention to content on GenAI-based HIS from social media was associated with lower perceived hallucination and privacy risks. Perceived hallucination risks, but not privacy risks, were negatively associated with attitudes toward GenAI and continued use intentions. Multigroup analyses revealed that machine heuristics moderated the relationships between social media attention, perceived hallucination risks, and attitudes. Theoretical and practical implications are discussed.
