A study by Carnegie Mellon, MIT, and Cornell found that roughly seven-minute conversations with Google Gemini reduced conspiracy beliefs more effectively than a static fact sheet in two experiments following the assassination attempt on Donald Trump and the murder of Charlie Kirk. Effects persisted weeks later, including lower agreement with general conspiracy narratives. GPT-4o screened participants; the models relied on curated fact bases since events occurred after training cutoffs. The authors note potential abuse risks.
No score is assigned. Sources and their independence are shown in the citation chain below.