Suk, J., & Zhang, Y. (2026). From mass media and social media to AI: A multilevel framework for understanding trust in generative AI. Journal of Broadcasting & Electronic Media, 1–24. https://doi.org/10.1080/08838151.2026.2694048
Summary
This article develops a conceptual framework for understanding public trust in generative AI by drawing on decades of communication research on trust in mass media and social media. Suk and Zhang argue that generative AI represents a new kind of information intermediary whose trust dynamics cannot be captured by any single level of analysis. They propose that trust in generative AI must be theorized across three interdependent levels — individual, institutional, and societal — and position the paper as a theoretical scaffold for future empirical work on how publics come to trust (or distrust) generative AI systems.
Key Contributions
- Proposes a multilevel conceptual framework (individual, institutional, societal) for studying trust in generative AI.
- Bridges the mass media trust and social media trust traditions with emerging AI scholarship, situating generative AI within a longer lineage of information intermediaries.
- Clarifies both continuities and distinctions between trust in legacy media, platforms, and generative AI, including new features such as generative outputs and human-AI interaction.
- Offers a roadmap and vocabulary for communication scholars to design empirical studies of trust in generative AI.
Methods
Theoretical and conceptual synthesis. The authors review prior scholarship on media trust across the mass media and social media eras, extract transferable insights, and integrate them into a new multilevel analytical framework applied to generative AI.
Findings
- Existing media trust frameworks are informative but insufficient on their own for generative AI, which introduces novel dynamics such as generated (rather than curated) content and interactive human-AI exchanges.
- Trust operates across interdependent levels: individual perceptions and use, institutional practices of AI developers and deployers, and broader societal contexts jointly shape trust outcomes.
- Treating generative AI as continuous with — but not reducible to — prior information intermediaries clarifies what is genuinely new about trust in these systems.
Connections
This piece provides a theoretical backbone for empirical work on how people perceive and evaluate generative AI, connecting closely to studies of user trust and interaction with LLMs such as Wang2025-zy and to work on AI as a news and information intermediary like Dierickx2026-tw and Hameleers2026-mc. Its emphasis on institutional and societal dynamics also resonates with critical and infrastructural accounts of AI’s role in the media ecosystem, such as Hepp2026-oi and Gillespie2026-aa.
Podcast
A research-radio episode discusses this paper: 🎧 MP3 · Spotify