Steven Rosenbaum’s book The Future of Truth, which explores AI’s impact on reality, unexpectedly included several synthetic quotes created by AI tools. Despite this, Rosenbaum insists AI remains indispensable in his writing process, emphasizing a future with cautious but continued AI use.
- AI-generated quotes discovered in Steven Rosenbaum’s book
- Author commits to refining fact-checking and continued AI use
- Raises concerns about verifying AI-assisted research accuracy
What happened
Steven Rosenbaum, journalist and author of The Future of Truth, discovered that several quotes in his new book were synthetic creations of AI tools rather than sourced from real interviews or texts. These included statements falsely attributed to figures like Kara Swisher and Lisa Feldman Barrett. The New York Times investigation brought these inaccuracies to light, leading Rosenbaum to implement a comprehensive citation audit with his editors to address the errors.
Rosenbaum used AI programs such as ChatGPT and Anthropic’s Claude to assist in research phases, including idea discovery and fact gathering, but insisted that the narrative and final writing were his own. The AI-generated research input was flagged in his notes and subjected to fact-checking and copy editing. Despite this layered review process, some fabricated quotes slipped through, highlighting new vulnerabilities in verifying AI-assisted content.
Why it matters
This incident underscores a growing challenge for authors and publishers as AI tools become more integrated into writing workflows. Traditional fact-checking methodologies rely on verifying directly sourced quotes and written materials. However, when AI synthesizes information—sometimes inventing plausible but false quotes—it complicates verification and raises the risk of publishing inaccuracies.
Simultaneously, the publishing industry faces cuts in editorial staff and fact-check teams, increasing reliance on AI tools while reducing human oversight. Rosenbaum’s experience illustrates how even high-profile works with editorial support are vulnerable to AI-induced errors, calling for stronger safeguards and more skepticism toward AI outputs in research.
What to watch next
Rosenbaum is moving forward with a thorough audit to correct his book and has expressed a commitment to being more cautious with AI-generated content. Observers will be watching how this audit influences future editions and whether it sets new standards for transparency in AI-assisted writing.
More broadly, the publishing industry must evolve its fact-checking and editorial processes to address AI’s unique risks, balancing AI’s creative benefits against the potential for misinformation. Stakeholders should monitor how tools develop and how guidance for AI use in journalism and authorship improves to ensure accuracy without sacrificing innovation.