AI Chatbots and Summaries Implant False Memories That Last a Week

Studies from MIT Media Lab and Georgetown University found generative AI chatbots produce three times more false memories than control conditions. A misleading AI summary cut accurate recall of a traffic sign from 83.6% to 44.8%, and telling participants the content was AI-generated did not remove the distortion.

• Chatbot condition produced 3x more false memories than control
• Misleading AI summary cut sign recall from 83.6% to 44.8%
• In MIT's witness test, 36.4% of responses were misled by the interaction
• False memories persisted one week with elevated confidence