Early this month, WIRED published an excerpt from Steve Rosenbaum’s hot new book. true futurewhich examines how artificial intelligence distorts people’s sense of reality. Shortly thereafter, the New York Times reported that the book contained no fewer than half a dozen fabrications or misquotes. Rosenbaum, who holds a master’s degree in “truth” from New York University, acknowledged in a statement that he had mistakenly included “a small number” and “inappropriate attributions or composites” of citations. Ironically, the veracity of a book about how AI affects truth is currently under intense scrutiny because of how the author used AI.
WIRED reviewed our 1,450-word excerpt after the Times story was published. Our fact-checking team reviewed it before publication and double-checked its quotes and facts for accuracy. However, WIRED’s AI-generated editorial policy prohibits the publication of AI-generated or AI-edited texts, and a reader email accusing the excerpt of being “blatantly written by AI” raised further questions about the extent to which Rosenbaum was using AI tools. in true futureIn the acknowledgments section, Rosenbaum writes that ChatGPT, Claude, NaturalReaders, ProWritingAid, and Grammarly “helped me refine and polish my presentation.” [his] idea. ” What on earth does that mean?
WIRED ran the excerpt through several AI detection services, including Pangram, GPTZero, and ZeroGPT. Each service suggested that it was likely or likely to have been generated by an AI. However, AI detection tools can be fallible and return inaccurate measurements. WIRED’s head of research emailed Rosenbaum directly, asking if and how AI was used to write the excerpt.
He replied: “Like many writers working today, I used AI tools for parts of the research and editorial development process for this book, including source discovery, brainstorming, structural feedback, and language refinement.” However, he emphasized that “the ideas, the reporting, the discussions, and the ultimate authorship are mine, and the excerpts in WIRED were not just generated by AI and published verbatim.” He urged WIRED editors to be careful about trusting AI detection tools, noting that false positives can occur.
At this point, a senior editor at WIRED asked me to look into this episode, since I’ve been covering AI missteps in various forms since 2024. My first step was to run the entire book through Pangram’s detection tool. (All AI detection tools have limitations and can give false positives, but Pangram is the current gold standard.) We found that 53 percent of the books were generated by AI, and a further 9 percent were recorded as potentially AI-assisted.
I called Rosenbaum and asked for more clarification on how he used AI to write the book and whether he disputed Pangram’s results. (BenBella Books, the colophon was published true futurea request for comment was not returned. Simon & Schuster, which sells Ben Bella’s books in the United States, declined to comment. )
Rosenbaum did not consider the accuracy of Pangram’s results. In fact, he didn’t want to talk about them at all. “I’m not going to participate in that conversation,” he said. “It’s like saying, Do you beat your wife? This is one of those accusations to which there is no answer.”
In return, he offered to broadly explain his editing process. He says he used AI tools as a search engine at the beginning of the writing process, helping him surface information in the more research-intensive sections of the book. To demonstrate how to do this, he asked ChatGPT to explain me and read out the results. AI Search more or less accurately described some of my previous stories, including my work on AI-generated “zombie media sites.”
