Trusting health information from large language models, online articles, and discussion forums: A comparison of perceived credibility

While the increasing use of Large Language Models (LLMs) as eHealth tools facilitates rapid information retrieval, it also carries the risk of credibility inflation for medically inaccurate content. The aim of the study was to examine the effects of the form of presentation of health information (LLM response vs. online news article vs. online forum) and its truthfulness on perceived credibility judgments. The study ( N = 405) employed a mixed-design experiment, with presentation format as a between-subjects factor and true vs false health information as a within-subjects factor. The results showed that although participants were generally able to distinguish between true and false information, the form of presentation significantly moderated this process with a medium effect size ( η 2 p = 0.07). Specifically, both LLM and online forum elevated credibility rating of false medical information. The findings have important implications for understanding cognitive mechanisms underlying credibility judgments in digital environments and for the design of safe and responsible eHealth tools.