When Narrative Correction is Delivered by AI: Testing the Effects of Narrative Correction and AI Fact-Checking on Debunking Health Misinformation

Narrative-based fact-checking has been proposed as a promising strategy to improve the effectiveness of corrective messages, as narratives can lower reactance. However, less is known about whether the source of narrative correction shapes its effectiveness. As social media platforms increasingly adopt AI-based fact-checking tools, it is essential to understand how fact-checking strategies function in the age of AI. This study conducted an online between-subjects experiment testing the effects of fact-checking message format (statistical vs. narrative) and fact-checking source (human vs. AI) on counterarguing against fact-checking, perceived credibility of misinformation, and belief in misinformation. Results showed that narrative fact-checking reduced both counterarguing and perceived credibility of misinformation compared to statistical fact-checking, but it did not significantly reduce belief in misinformation. Furthermore, narrative fact-checking was more effective at reducing perceived credibility when delivered by a human rather than AI, whereas statistical fact-checking was more effective when delivered by AI rather than a human. These findings offer implications for social media platforms in tailoring fact-checking strategies based on message format and source.