Journalist and author Steven Rosenbaum has begun a full citation audit of his new book, The Future of Truth: How AI Reshapes Reality, after an investigation by a news organization identified improperly attributed or synthetic quotes linked to his use of AI tools during research. Six problematic citations were identified in the book, including three synthetic quotes with no apparent source, out of 285 outside citations.

Rosenbaum is working with editors on the audit to correct future editions. "I learned a lesson and will be more suspicious and reticent to trust AI outputs going forward," he said, adding that he would not pause AI use and return to manual writing tools like Microsoft Word.

Tech reporter Kara Swisher said a quote attributed to her in the book is one she never said. Professor Lisa Feldman Barrett said quotes attributed to her do not appear in her book and are wrong.

Describing his research process, Rosenbaum said he used AI tools such as OpenAI's ChatGPT and Anthropic's Claude to surface ideas, locate articles, summarize themes, and identify people or papers to look into. He said AI was never used to write the book's reporting, narrative structure, interviews, arguments, or conclusions, and that the book includes chapters based on interviews he transcribed himself as well as research-based chapters in which facts are pulled and knit into a narrative.

He said any nuggets mined by AI tools were tagged with a warning in his notes, and those tagged notes were passed on to a fact-checker and two copy editors provided by the publisher. "I did not set out to fabricate anything," he said, describing how AI-generated information looked authoritative and made its way too far downstream before being caught. He said publishers will need new verification workflows for AI-era research, including mandatory source tracing for quotations, better provenance tracking, clearer standards around AI-assisted research, and AI tools that audit citations against primary materials.

Similar incidents have occurred elsewhere. The Chicago Sun-Times last year printed an advertorial summer reading list containing non-existent books generated by AI. A correction was issued after published quotes attributed to Conservative leader Pierre Poilievre turned out to be an AI-generated summary of his views. Publications including Wired and Business Insider removed multiple articles attributed to "Margaux Blanchard" that appeared to be AI-generated. Ars Technica retracted an article after a former reporter used an AI tool to extract quotes from a blog post but the tool generated fabricated versions. Pre-print clearinghouse arXiv implemented a zero-tolerance ban policy on papers with hallucinated citations.

Rosenbaum said the irony of incorporating AI-generated falsehoods into a book on AI's reality-skewing effects is uncomfortable, and that failures in his own book more vividly demonstrate the problem of pervasive and persuasive AI systems than any abstract argument. "Sometimes negligence is just negligence and not every failure is instructive," he said.