A Book About AI and Truth Was Caught Using Quotes That AI Made Up

A nonfiction book built around a warning about AI and misinformation turned out to have its own AI misinformation problem. The story, first reported by The New York Times, is a blunt illustration of the exact research risk it was published to describe.

What actually happened

Filmmaker and author Steven Rosenbaum published "The Future of Truth: How AI Reshapes Reality" through Matt Holt Books in May, pitched as a serious look at how generative AI is eroding our shared sense of what is real. It picked up favorable early coverage. Then a Times reporter checking facts for a piece on the book started running down its citations and found more than half a dozen quotes attributed to real people, including Northeastern professor Lisa Feldman Barrett, that those people never said.

Rosenbaum acknowledged the problem when the Times asked him about it. He said he had used AI chatbots as research tools while writing the book, noted in his own files which claims came from AI, and passed those claims along to a fact-checker working for the publisher. The hallucinated quotes made it through that process anyway and into the finished book. Rosenbaum says he takes full responsibility for the errors and is now working with his editors to review the text and correct the affected passages, with fixes planned for future editions.

Why this matters if you write with AI

This is not a story about a careless author skipping fact-checking. A fact-checker was involved, and the errors still shipped, because the underlying problem with AI hallucination is that fabricated quotes can read exactly like real ones. There is no visual tell. A chatbot will invent a plausible sentence and attribute it to a real, credentialed person with the same confident tone it uses for a sentence it got right.

That is the lesson for any nonfiction writer using AI to speed up research: treating an AI tool's output as a lead to verify is fine, treating it as a citation you can drop straight into a manuscript is how a book ends up correcting itself in public. The more specific and quotable a claim sounds, the more it deserves an independent source before it goes anywhere near your draft.

How to keep AI research out of your final manuscript unchecked

A few habits keep this exact problem from reaching your own readers:

• Trace every quote and statistic an AI tool gives you back to a primary source before it goes in your manuscript. Our guide on how to use AI for research when writing a book covers how to use AI for leads without treating its output as fact.
• Build a habit of flagging AI-sourced claims in your own notes so a fact-checker, or you on a second pass, knows exactly what still needs verification. How to avoid low-quality AI writing outputs covers the broader version of this discipline.
• Decide what you will disclose about your research process before a reader or reporter asks. AI writing ethics: what authors should know walks through the questions worth answering in advance.
• Rewrite and re-verify any AI-assisted section until you would stand behind every claim in it personally. How to edit and humanize AI-generated book content is a good starting point.

PageWriter Studio keeps your research and your drafts in one place as you write, so it is easier to see at a glance what still needs a source before a chapter is finished, instead of discovering the gap after the book is out. If you want a writing workflow built around staying accurate as well as fast, you can start a free trial today.