The irony here is almost too on the nose: a book written specifically to examine whether AI can be trusted with the truth turns out to contain quotes that an AI invented wholesale.
This is a textbook case of AI hallucination - the tendency of large language models to generate text that sounds authoritative and plausible while being entirely made up. The models aren't lying in any deliberate sense; they're pattern-matching machines that produce fluent, confident prose whether or not the underlying claim is real. They have no internal alarm that fires when they're fabricating a quote versus recalling one accurately.
The problem is especially sharp with direct quotations and citations. A model asked to find what a specific expert said on a topic will sometimes produce a quote that sounds exactly like that person, attributed correctly, with plausible context - and none of it will exist in any source. Researchers have documented this failure mode repeatedly, and it has caught out lawyers, academics, and journalists who used AI for research without verifying the outputs.
For authors and publishers using AI tools to assist with research or drafting, the practical rule is blunt: every direct quote, every attributed statement, every citation needs to be checked against the original source before it goes to print. Searching for the quoted text verbatim is the minimum standard. If the source can't be located, the quote doesn't go in.
The deeper problem is that AI-assisted writing can produce text that feels thoroughly researched. That confidence is baked into how these models write - it's not a feature you can turn off. Anyone using ChatGPT, Claude, or similar tools to help with a book manuscript needs to treat every generated citation as a hypothesis to verify, not a fact to copy.