Equally Credible? Examining Credibility Perceptions and Evaluation Processes of AI- and Human-Generated Information
As large language models are now capable of producing extremely humanlike content, it is crucial to understand how people assess the credibility of such information. Using Lee’s integrated theory of human-machine communication as a theoretical backbone, this study aims to examine whether information attributed to AI is considered equally credible as information attributed to humans and to identify the credibility cues used to evaluate each type of information. In an experiment, participants evaluated the credibility of news articles or social media posts labeled as AI- or human-authored as mostly equal, although human-authored social media posts were perceived as slightly more credible than AI-authored social media posts. Participants also evaluated AI- and human-generated information similarly in terms of how many and which credibility cues they relied on to make their evaluations. Importantly, prior AI attitudes did not affect participants’ credibility evaluations. The results are interpreted in the context of rapidly changing AI technologies and public experience with AI-generated content.
Ptaszek, G., & Metzger, M. (2026). Equally Credible? Examining Credibility Perceptions and Evaluation Processes of AI- and Human-Generated Information. International Journal of Communication, 20, 1972–1990. https://doi.org/10.65476/karyzt10