Large language models reshaped news consumption patterns starting in August 2024, but publishers who blocked AI bots experienced worse outcomes than those who didn't, according to research published December 31, 2025, by Rutgers Business School and The Wharton School.