Listen: Readers spot AI writing 64% of the time: the first 1,636 answers from the AI vs Human test

Can you spot the difference between a journalist's writing and an artificial intelligence rewrite? In a recent test, readers were shown pairs of short passages cut from the middle of news stories and asked to identify the machine-generated version. Out of more than sixteen hundred answers, humans got it right about sixty-four percent of the time. That is significantly better than a fifty-fifty coin toss. But the average hides a massive divide. Some players achieved perfect scores, while others struggled to beat random guessing even after dozens of tries. The exact same text that one person reads with near-perfect accuracy leaves another completely stumped. The single biggest factor in success was patience. Speed kills accuracy. Wrong answers were decided in about half the time of correct ones. When players rushed and answered in under five seconds, they were barely better than a coin flip, scoring just fifty-five percent. But when they took their time, spending forty seconds or more to analyze the text, their accuracy jumped to seventy-five percent. It turns out we can spot the machine, but only if we slow down and actually read.