When the worried are right about AI

The real close

I’ve spent this series talking you down from fears that were aimed at the wrong target. Some of the worry is correct, though, and pretending otherwise would make everything else I said cheaper. The authors who get hurt by AI are the ones who ignored the legitimate fears, the situations where the worry was accurate and they walked into the trap anyway.

Three worries are correct, and I won’t wave them off with the rest. I’d give up some of the comfort this series offered before I’d let you walk into one of these carrying my reassurance.

Get these three right and everything else I’ve said applies to your book. Get them wrong and no amount of reassurance will save it.

When will AI really take the work?

If the work is commodity content with no voice in it, the kind of writing where any competent version is interchangeable with any other, that work is going away, and no amount of optimism changes it. Product descriptions, filler blog posts, generic summaries, boilerplate explainers: AI does all of those well enough and free, and the market has already noticed. Rates for that kind of writing have collapsed, and the freelancers whose offer was “I’ll produce clean copy at scale” are scrambling.

If your plan depended on being paid for writing that carried no fingerprint, the worry is correct and the plan needs to change. The escape is moving up into the kind of writing where voice and judgment matter. That work is structurally protected. AI raised the value of it by automating everything below it.

I don’t have much sympathy for the content mills that built businesses on interchangeable copy.

I have plenty for the freelancers they paid next to nothing to produce it. Those writers were underpaid for work nobody valued, and now the floor is gone. If that’s you, the climb into voice-driven work is hard, and it’s the one road still open.

I wrote at length about where AI is and isn’t coming for writing work, and the dividing line is exactly this one. On one side, the work is gone. On the other, the work got more valuable. There’s not much in between anymore.

When will AI embarrass you?

Turn AI loose with nobody checking its output, and it’ll eventually humiliate you in public. The machine states invented facts in the same confident tone it uses for true ones, and it does this on every model, more than once in a long document. An author who publishes that unchecked, under their own name, is being careless with a tool that was always going to do this.

The lawyer fired for a brief full of made-up case citations wasn’t the exception.

He was the example. The journalist whose AI summary invented quotes was the example. The book whose factual errors got dragged on social media and pulled from the platform was the example. Every one of those incidents started with a person who trusted the machine to be right and never checked. The fear of getting burned by a hallucination is accurate, and the fix isn’t better prompting.

The fix is verifying every specific claim by hand before it ships. The hallucination survival guide covers what to verify and how, in detail. If you publish anything AI touched without doing that verification, you’re betting your name on a machine that will eventually let you down in print.

This one is inexcusable, because it’s so easy to prevent. Checking a citation takes minutes. Skipping it can cost an author the trust their book was supposed to build, and no correction notice brings that back.

When will AI make you sound like everyone else?

Hand the machine your voice and it’ll flatten you into the average, exactly as you feared.

This worry is correct whenever you cross the one line that matters. Let AI organize your research and you’re fine. Let it write the narrative a reader will attribute to you, and you’ve traded the only thing that made the book yours for a speed advantage you didn’t need.

The authors who lose their voice to AI don’t lose it all at once. They lose it gradually, by letting the machine write more and more of the parts they were supposed to write themselves. First a section, then a chapter, then a whole stretch where the author was tired and the machine was available and the prose was good enough to ship.

By the time the manuscript is done, half of it is in the machine’s voice and half in theirs, and the reader feels the seams even if they can’t name them. The deeper piece on losing your voice to AI covers how this happens and how to prevent it. The short version: draw the line at the narrative, and keep a human on every word of it.

What to do about each

The same rule answers all three. Point AI at work that has no voice in it and that a person can check afterward: the research, the structure, the dull connective sections. Keep a person on the voice, the judgment and the stories. Verify every fact by hand before it goes out. I give authors that rule no matter who they are, because the line doesn’t move for anybody.

That rule is the only arrangement that produces a book worth publishing now that the machine is in the room. The cornerstone piece on whether AI can write your book walks through where to draw that line in more detail, and the rule has held across every project I’ve run since AI became serious enough to use.

Why should authors take these AI fears seriously?

Any guide that tells you everything is fine is selling you something, and I’m not interested in writing one. The worries above are real, and you’re right to carry them. They’re also avoidable. The authors who get hurt are the ones who shrugged those fears off, and the ones who do well took them seriously enough to draw a line and stay on the right side of it.

So worry less about detection, about being too late, and about whether a human is still worth paying for. Keep all of your worry for the three things above. I’d be failing you if I let the reassurance cover those, because they’re where AI hurts the authors who trusted it too far, and every one of them is avoidable if you do the work.

From the author

The Death of Thinking
Some AI fears deserve action, and the loss of judgment tops my list. The Death of Thinking covers how that loss builds up through decisions that each looked reasonable.

Frequently Asked Questions