1 article
Experiments show 70%+ of participants accepted verifiably wrong AI answers without questionâeven when the errors were glaring.