In this video, Julie Broad (founder of Book Launchers) warns nonfiction authors about the hidden dangers of using AI for research and writing. She explains that AI often generates confident but inaccurate statements, leading to a dramatic increase—up to four times—in fact‑checking time for manuscripts that used AI. Through concrete examples (a mountain named after a dog, fabricated journal articles, mis‑matched citations, and near‑plagiarism of chakra definitions), she shows how AI can produce fabricated facts, fake sources, and subtle definition shifts that easily slip past authors. Julie stresses that authors must personally open and verify every cited source, keep simple notes of URLs or titles, and never rely on AI as a final fact‑checking authority. While she embraces AI for brainstorming, outlining, summarizing, interview prep, and marketing ideas, she cautions against using it for medical, legal, historical, or statistical claims. The video concludes with a call to use professional human fact‑checkers and offers Book Launchers services and community resources to help authors maintain credibility.
Chapters
The scariest thing about AI isn't that it lies. It's that it lies so confidently and so well, you probably have believed it. And if you're using AI to help research or write your book, there's something you need to know before you hit publish. So if you wanna make sure you're not publishing lies, this video is for you. Hi, I'm Julie the Book Broad. I'm the founder of Book Launchers and your host here at Book Launchers. Booklaunchers.tv. Book Launchers is your first choice for all things related to writing, publishing, and promoting a nonfiction book. I am a really big fan of AI. I mean, we've built an entire software program, our Author Launch Kit, with AI underneath. So I am not an anti-AI person, but I have to tell you something wild. So at Book Launchers, we have a team of editors, and one of the editing steps that we take after content edit and copy edit is a fact check. And our full-time fact checker has flagged something really interesting after I questioned all the time that was going into fact checking this year that wasn't going into fact checks in previous years. And in fact, our time spent fact checking hasn't just doubled, it's nearly quadrupled on a lot of projects. And guess what? The projects where it's quadrupled, the author has used AI at some point in their journey. And the challenge is AI does not know what's true. It predicts what sounds true. And that difference is becoming a massive problem for authors. Take this example we ran into. The manuscript stated that Mount Ballyhoo was named after explorer Edmund J. Ballyhoo. Sounded right. But our fact checker double-checked it. And guess what? The mountain was actually named after Jack London's dog. Yeah, it sounded so good, the author never questioned it. But imagine if the fact checker hadn't, or imagine if that was a more serious issue that had been written wrong. So before you start hitting your ChatGPT for your research, remember, it's not a search engine. AI tools are generators, not necessarily truth generators either. They predict likely language. They synthesize patterns. They're designed to sound coherent, but sounding coherent is not the same as being accurate. It's one of the things I always flag for people. I'm like, "Used AI." They're like, "Just to edit?" I'm like, "Yeah, it flattened out the humanity out of your book." But the danger isn't obvious nonsense. The danger is information that's 92% correct, and that 8% is where you're in trouble because it seems so true. And that's the issue we're running into and why it's taking us 4 times as long to edit manuscripts that used AI, because we're seeing wrong statistics in the wrong context, invented supporting claims, Fabricated facts, fake experts, fake studies. One of the books we worked on had most of the facts true, but it added a few things in for good measure. For example, it said it benefits brain, liver, and kidneys, but the sources only discussed benefits to the brain and kidneys. The liver was never mentioned in the sources. Sometimes we'll get a source that is a real source, but the actual material it's supposed to talk about Nowhere to be found in the source. So asking AI to give you a source might feel like you're getting protected, but you're actually not covering your butt because you have to look at what that source is and make sure it actually says what it is supposed to say. When you publish a book, people assume you checked your facts. Your credibility is on the line. Now imagine you have a made-up fact in your book and you don't realize it and your book takes off. And suddenly you find podcasts citing your book, media referencing your book, clients are trusting your expertise, readers are repeating your claims, and then somebody discovers that one of those claims is not true. I'm a fraud. One fake thing can poison trust. And of course, you never know the sources that were used to produce the data. Were they trustworthy? And also, are you unknowingly plagiarizing someone else's work? Authors think plagiarism means copy and pasting entire paragraphs or even just one lines, and they recognize that as, you know, plagiarism or something that needs sourcing, but they don't always source when something came from AI. And AI often creates near copy content that feels original enough to slip through. But for example, in response to the prompt, describe the root chakra, it says that root chakra is associated with safety, stability, grounding, and feeling secure in the world. While a popular online chakra resource uses the exact phrasing connected to safety, stability, and feeling grounded in the world. The AI response has just swapped a few words. If you take what AI says, you are plagiarizing. Also, there are some subtle things worth looking at, like if you ask it to do a definition. We tested this with the word advocate or advocate. How do you advocate? Advocate. Advocate? We tested it with the word advocate. Merriam-Webster says, "To support or argue for." AI says, "To support or argue in favor of." That's dangerously close, but also different. So the scariest thing, and why everyone who let AI touch their book absolutely should be working with a professional human fact-checker, well, the crazy thing is we're seeing this over and over again. Imagine citing a source in your book only to discover the source never existed. Yeah. And it's even worse than that because some of it is true, usually just enough to be believable. Like one of our authors sourced a real journal, but the article was fake. There was no such article. Or as I already mentioned, we had a real article that was sourced, but there was nothing in that article related to the material in the book. We've also seen mixed up author names. And of course, I'm sure you heard the crazy story last summer of the bestseller list with a book by a famous author. Author was real, but he'd never written the book. This is so common with AI citation cleanup. And this is why my team begged me to make this video to let all authors know that you should never cite a source you have not personally opened and reviewed yourself. Always aim to cite your actual original source for your research. This is a lot easier to do if you take notes on your sources as you do the research. You don't have to write down it properly formatted. Just jot down a source's URL or a title. You don't have to create the full citation. Just note where you heard this or read this or learned this. And then never cite a source without reviewing it first. And don't trust that the content of a source AI has cited matches AI's write-up. Always find a claim in a source before you cite it. If using AI to edit and clean up your manuscript, craft your prompts carefully and carefully review all outputs. Again, we use AI constantly at Book Launchers. We like it as an intern though, not as our senior editor. It is not about whether you used AI in your book at all. It's really how and for what. And I mean, I love it for brainstorming, outlining, even summarizing. Idea generation, interview prep, drafting possibilities, brainstorming content ideas for marketing, identifying gaps, reader research. But it is not going to be a great final fact-check authority. I don't recommend it for editing. Like I said, it flattens things out. It introduces errors. It makes it sound good, but it's not human anymore. Medical legal claims, Dangerous. Historical accuracy? Dangerous. Statistics without verification? Well, hopefully you got the gist from this video. AI can accelerate your thinking, but don't let it replace your judgment. Take time to review everything you're saying in your book and make sure it's true. In the next decade, being deeply human is going to matter more, and being trusted is going to be everything. Your stories, your lived experience, your judgment, your discernment, That's the edge AI cannot replicate. Now, of course, if you want a team on your side to help you produce the absolute best quality book, including our awesome team of fact-checkers, well, go ahead, set a call up at booklaunchers.com/apply. If you want help building that nonfiction book that actually builds authority instead of damaging it, you can also join our free community over at booklaunchers.com/school. I'll see you there.
