How writing mistakes happen, and how to catch them (even if AI helped you write)

I’ve written two books, coauthored three more, and ghostwritten four others. I’ve also done developmental edits on three dozen.

I’m proud to say that there are very few mistakes in any books I’ve worked on. That’s not because I don’t make mistakes — I make plenty. But I don’t trust myself to be completely accurate, because sometimes stuff gets messed up in the writing process and needs checking.

And when I do developmental edits, I certainly don’t trust what I’m reading. I’m being paid not to trust it. I try to notice every mistake and fix it, or at least tell the author to fix it.

What I check

Of course I check what the document is readable and logical and the text is grammatical. But I also check these things for accuracy:

• If there is a number, is there a verifiable source that shows the number is accurate? Is it footnoted?
• If you do math, did you get the answer right?
• If you say an event took place at a given time and place, is there evidence that it actually did?
• If there is a quote from a public figure, did they actually say what the text says they said? Where’s the source? (No, George Bernard Shaw did not say “The single biggest problem in communication is the illusion that it has taken place.” And Maya Angelou did not originate the saying “They may forget what you said, but they will never forget how you made them feel.”)
• If there is a direct quote from an interview you did, did the person you interviewed actually say the words you have attributed to them?

In 20 years as an author, I’ve made these mistakes thousands of times. That’s not the point. The point is, I caught them and fixed them.

After I finish with a manuscript, it goes to a copy editor. Often, the copy editor will find mistakes I missed. But not many. I take pride in the fact that by the time it gets to the copy editor, there should be hardly anything else to catch.

AI means you must check even more diligently

AI-generated text includes few grammatical errors. But it has a high likelihood of including factual errors.

AI makes math mistakes.

AI gets facts and dates wrong.

AI misattributes ideas.

It makes up quotes.

It attributes quotes to the wrong people.

Why? Because accuracy is hard. AI can read a passage and misinterpret it. It can read a Reddit post and be confused about whether it’s a definitive source. It can make a transcript that includes errors. It can read that transcript and mis-cite it, introducing additional errors.

Like a well-dressed, fast-talking salesperson, you can’t trust it. It creates text that looks awesome, but that doesn’t make it right.

Just because it’s faster to generate prose doesn’t make that prose bulletproof and accurate.

Checking is slow and tedious. But it’s even more important if you’re using AI to generate content, because you don’t even have the experience of making the mistake. You’re more prone to miss it.

Don’t count on the copy editor to bail you out. Checking facts and quotes is your responsibility.

Good authors aren’t just inspiring. They’re accurate. And like everything else in writing a book, that’s work for which there are no real shortcuts.