This article digs into how journalist Steven Rosenbaum used AI in his research for The Future of Truth. It also looks at why the New York Times uncovered AI-generated errors in his book—despite Rosenbaum’s defense of the technology as a helpful tool.
We’ll break down what went wrong, the exact failures the Times found, and what this all means for verifying sources and the future of AI in journalism.
What happened and why it matters
The New York Times found six problematic citations in Rosenbaum’s book out of 285 outside sources. Among these, three quotes were simply made up, with no traceable source.
The Times’ investigation set off a formal citation audit. This kicked off a bigger conversation about how Rosenbaum—and others—use AI in their research.
Rosenbaum says AI surfaced ideas, found articles, summarized themes, and flagged possible sources, but he handled the reporting, interviews, and the book’s narrative.
He also points out that he passed tagged AI-generated notes to a fact-checker and two copy editors. Still, the team didn’t catch every error.
AI-assisted research: benefits and vulnerabilities
Rosenbaum calls AI “magical” for connecting ideas and making research faster. He lists perks like pulling text in seconds, finding related articles, and surfacing themes from massive piles of data.
But it’s not all smooth sailing. He recalls times when AI just made stuff up—even when he gave clear instructions. Sometimes, the AI’s confident but wrong responses looked real enough to fool someone.
The big issue? AI can quickly point out patterns and sources, but it might also invent quotes or mix up attributions if humans don’t double-check everything.
Rosenbaum admits that traditional fact-checking just isn’t built for this. There’s a real need for new standards and better guardrails.
- Benefits: faster brainstorming, quick source discovery, and rapid theme summaries.
- Risks: fake quotes, missing or incorrect sources, and too much trust in AI output.
Rethinking newsroom verification
The Rosenbaum situation highlights a real paradox. The same tools that make research faster can also wreck credibility if no one’s watching carefully enough.
Some critics can’t help but note the irony: a book about AI gets tripped up by AI-generated errors. It’s not just a warning—it’s a wake-up call.
People pushing for reform say newsrooms have to update their verification workflows to keep up with AI. That means tracing every source, tracking where AI-generated notes come from, and keeping clear audit trails that show how algorithms shaped quotes or references.
Practical guardrails and standards
- Set up mandatory source tracing for all AI-suggested references. Always give clear credit to the original materials.
- Track provenance for AI-generated notes. Make it obvious when a piece of text or a clue came from an AI prompt instead of human research.
- Build AI-specific verification standards that call for independent checks on AI-generated quotes or claims. Even if something sounds convincing, someone still needs to confirm it.
- Use a layered fact-checking workflow. Start with an AI-assisted pre-check, then bring in human editors. Document all edits or swaps clearly.
- Put money into AI auditing tools so you can review citations, quotes, and attributions for accuracy and where they came from.
Rosenbaum still believes AI can transform research, even after some bumps along the way. He points out that the speed—like how quickly you can pull out key info or map ideas—makes it worth the effort, as long as you build in better safeguards.
As AI spreads, readers really do deserve to know how algorithms shaped what they’re reading. Publishers need to step up with stronger verification standards if they want to keep trust intact.
Bottom line: AI’s not going anywhere in journalism. If newsrooms and writers want to use it well, they’ll need solid systems for provenance, traceability, and verification. That’s the only way to catch AI slip-ups before they hit the public—and keep that sense of discovery alive without losing accuracy.
Here is the source article for this story: AI put “synthetic quotes” in his book. But this author wants to keep using it.