Research · Georgetown
Georgetown research finds misleading AI summaries can distort human memory
A Georgetown-led study found that inaccurate AI-generated summaries reduced participants' ability to recall details from an observed event.
Confirmed facts
Researchers from Georgetown University and the University of Washington studied errors in summaries generated by ChatGPT and Gemini and tested how inaccurate summaries affected participants’ memories of an animated event. Across the model outputs they reviewed, summaries omitted 51.6% of central details on average.
Among 331 participants, 83.6% of those who read an accurate summary correctly recalled whether a car approached a stop or yield sign. Accuracy fell to 44.8% after participants read a misleading summary. Telling participants that the summary came from AI did not eliminate the effect.
Evidence boundary
The study used animated videos and a controlled experiment. It does not establish that every AI summary distorts memory or measure performance in operational settings such as policing or medicine. The authors identify those settings as consequential areas for further study.