One vs. Many: Comprehending Accurate Information from Multiple Erroneous and Inconsistent AI Generations