Two documents were uploaded. The requirement was to compare the information in Document 1 with that in Document 2. After correctly extracting the first 20 pieces of information, the user began to fabricate additional content arbitrarily. Despite multiple reminders, the user refused to make any corrections—such behavior is actually less reliable than using free AI tools.