Readers thought they preferred human authors. Then researchers put them to the test. The result? AI-written stories got the higher scores. How is that possible?