Writing in the time of AI

Where do we even go from here

Okay so.

A couple of things before I get going. I’m going to use “AI” because that’s what folks are calling it, even though we know it’s not intelligent, not in the way of sci-fi books and movies. The specific flavor I’m talking about here is LLMs and other text generators that people use for search, conversation, and email writing. And sometimes, apparently, story writing. (Here are two articles talking about the way writers are using — or not using — AI in their own work. This one is survey results on BookBub, and this one is survey results from Australian writers. Note the sample sizes are both pretty small, ~1200 and ~400 respectively. (Update: a friend told me that’s actually a good sample size!)

Additionally, I’m not going to name books. You’ll figure out some, I’m sure, but please don’t name them in the comments. If/when things come out, there will be better sources than me. I’m not your source. I’m not here for the drama. Perhaps you’re aware that there’s discourse going on right now, but please understand I’m not speaking about that specifically.

I’m talking about how we, as authors and readers, might think about the future of books, writing, and our relationships with one another in a world with this type of AI.

---

Like a lot of authors, I spend too much time thinking about AI in the publishing industry, how it’s being used, and how it will affect our futures. You know I love writing, and it’s baffling to me that anyone would want to give up the absolute joy of getting to know characters, exploring new worlds, or fiiiinally figuring out something huge about a story. There’s nothing like that sense of victory after moaning for days that I’m the worst writer ever, I had a good run, now it’s time for me to go lie down in a ditch. NOT TODAY, DITCH. NOT TODAY.

By comparison, using AI to “write” seems like a very flat, bland experience.

And let’s not forget the facts that it’s trained on stolen work, devastating for the environment, and faces a ton of questions about copyright.

Ahem.

I have a lot of thoughts . . . and in no specific order:

---

I’m worried about the ways we try to figure out if something is AI.

Look, I’m not even talking about the clickbait posts of “if it has an em-dash/semicolon/serial comma, that means it’s AI!” We all know those are completely ridiculous, intended to make people angry enough to engage. Punctuation is not an indication of whether AI was used. Since it was trained on books, and books use punctuation, well . . . guess what!

So let’s move on to those “AI checkers.” And then move right on past them, because they’re famously inaccurate. There’s a pretty laughable example of them declaring that the United States Constitution is written by AI! They don’t work. And my gosh. The irony of using AI to check for AI.

Also, it shouldn’t need to be said, but . . . don’t feed someone’s work into an AI. Even to check that it wasn’t generated by AI.

Okay, so what’s another method? Well, recently I saw someone dissecting lines from a book they thought might be AI, using what they knew about how LLMs work to try to figure it out. And while I really appreciate that this method involved actual thought and consideration, I found myself disagreeing with them. Why?

Because they kept saying, “This doesn’t make sense,” and “A human wouldn’t write it like this.” And, uh, a human absolutely would.

Okay, this isn’t to make anyone feel bad, but bad writing exists. Boring writing exists. Weird writing exists. And sometimes it gets published. Honestly, I have looked at my own writing and wondered what the heck I was even thinking when I wrote it. Was I thinking? BRB going to check if that ditch is still available.

Confusing sentences are normal, human errors. We always hope they get fixed before printing, but . . . sometimes they’re published like that. Sorry. We’re human.

And hey, did you see this piece that includes two samples of writing, one written by a human and one written by a LLM? (Gift link.) It’s a little bit of a Turing test, though you’re reading already knowing that one of them is not authored by a human. That definitely affected the way I read both pieces.

Take a second and go read them. Let me know if you were correct about which one was human and which wasn’t. I correctly identified them, but I don’t think I would have identified the AI one as AI if I hadn’t gone in with the knowledge that one of them was.

So, without someone admitting to it, is there a real way to tell if a book is AI? Not to my knowledge. I mean, unless they leave in the prompts. You can be pretty sure then.

---

How . . . does this work socially?

At some point, it’s probably going to happen. Someone who used AI to generate their book is going to get a traditional deal. Maybe it will be an open secret. Maybe it won’t be a secret at all.

I’m imagining being on a panel with this person. Seeing them at a book festival. Trying to be professional when I know their “work” was only possible because human-authored books were part of the LLM training material. My work. My friends’ and colleagues’ work. And none of us consented to that. None of us were compensated, but now that theft, that violation, is financially benefiting some other person — someone who wants to be taken seriously as an author, even though they didn’t write their book.¹

I can see a future where authors have to make hard choices. If I’m invited to a festival, but I find out they’re also hosting someone who uses AI, am I willing to attend? Am I willing to lend my credibility to that festival, once they’ve shown that AI is not a line for them?

What about bookstores?² Libraries? Other authors who blurb/promote AI books? You can see how this becomes socially complicated very quickly.

I don’t want to have to make these choices, but I fear one day I will.

---

So what do we do?

There’s a lot of legal stuff going on behind the scenes. Cross your fingers. Call your reps to support regulations. (Senators specifically should be told to push back against the provision in the budget bill that prevents states from regulating AI for the next ten years.) Demand transparency where AI is used — in publishing and in regular life.

When you get a contract, make sure there’s language that says AI can’t be used to write or edit the book, or create the cover or marketing materials. Make sure your publisher can’t sell your book as training data. (This is something agents and authors have been working on for years, so let’s keep it up!)

But mostly: write your next book.

Because the way I see it, writing is our resistance. In a world that’s telling us human art isn’t needed anymore, because anything can be generated, continuing to create is a radical act.

Write with your whole self. AI-generated text is only going to become more technically proficient as the technology improves, so we need to write unique, engaging stories that are deep and memorable. Surprising, delightful, and thoughtful. The depth and nuance of our stories will be what elevates us. The authentic human voice will be more powerful than ever, so use yours!

I do think there are changes ahead, things we end up doing (or feeling like we have to do) to “prove” we’re just using our own little brains to write books. Authors can’t really do progress videos like visual artists can — and even those can be faked³ — but I’m sure things will pop up. Publishing schedules might slow down.⁴ There might be paper trails with drafts sent to friends and agents. I don’t know. People are creative, and tests will be issued in all kinds of ways. They probably won’t be fair.

And frankly, no one should have to prove they didn’t use AI every time they share a new thing.

But also, don’t use AI. If you’re a writer, don’t use it for your cover or marketing. Don’t use it for brainstorming or audiobooks. Don’t use it for any part of the process. As I said in my last newsletter about AI, make sure you’re not encouraging the exploitation of other artists. Don’t participate in it.

Not only is using it a betrayal of your colleagues, but it will be very difficult for people to believe if you say you use it there but not here. Or if you used it to write that book but not this book. Why would anyone believe you’d changed, especially if there’s a huge financial incentive to just . . . lie?

We should also be wary of making willy nilly allegations. We don’t want to normalize accusing people of using AI just to damage their reputation, you know? That will certainly undermine other allegations that have more truth behind them. It will hurt the cause. So stay skeptical.

And if something does come out as being AI, don’t rush to buy a copy and read it for yourself. There’s a good chance you won’t be able to tell anyway (again, unless they leave in the prompts . . .), and you’ll just be playing into the trick. The “authors” of AI books don’t care about you or your reading experience; they want your money. Don’t give it to them. And if it’s traditionally published, definitely don’t buy it. Publishers are for-profit businesses and they respond to sales. Vote with your dollars, as they say.

And no, this is not permission to pirate. Just starve them of oxygen.

---

The last thought I want to leave you with today . . . yeah. It stinks. And there are a lot of reasons to be upset and afraid. Most of us didn’t get into writing to get rich, but because we love it. (It is nice to be able to buy food though.)

But also, I’m seeing a lot of pushback on the idea of AI-generated books. While there are people who don’t care (cannot relate!), there are many who do. Readers want to read books by humans. They want to support human authors. And they don’t want to feel duped by picking up a book they thought a person wrote . . . only to realize it wasn’t a person.

There’s a reason why so many people who use AI try to hide it.

They know it’s unethical. They know they’re not doing the work. And they know it comes off as grifty — like they’re just trying to make a quick buck by tricking people into buying a thing they wouldn’t have otherwise.

So be your most human self. Keep making your art.

Boop.

Two other things:

The audiobook for BYE FOREVER, I GUESS is 50% for a few more days. Use this promo link. Think You’ve Got Mail but with texting, gaming, and knitting.

And Nashae Jones and I will be visiting a few Virginia Barnes and Noble stores this summer. You’re invited!

• June 14 at 1pm — Richmond, VA (Chesterfield Town Center)

• June 18 at 3pm — McLean, VA (Tyson’s Corner)

• June 27 at 1pm — Harrisonburg, VA

• July 5 at 1pm — Glen Allen, VA

• July 13 at 2pm — Fairfax, VA (Fair Lakes Promenade)

Will we see you at any of these events? Hope so!

1. Or have it ghostwritten, which is what normal people do when they want a book but know they don’t have the skill.

2. I imagine bookstores and so on will be forced to make difficult choices, too, if something comes out as AI. If it’s a big book that will sell well, do they just not stock it and lose the money it could bring in? Or do they stock it and risk upsetting their relationships with other authors? Big retailers won’t care as much — Amazon certainly won’t — but indie bookstores? Either way could be damaging for them. (I know what I hope they choose, but I’m not the boss of them.)

3. I thought about linking a source here, but the explainer video I saw is just a regular person on Threads and I don’t want to share without permission. So I did a little search for “how to fake a progress art video” and this immediately popped up. You understand why I’m not clicking on it and ruining my algo. That feed is fine-tuned for influencer drama, science, and knitting!

4. I’m not saying this is fair, because there are absolutely writers who can write that fast. But I suspect speed will be suspicious for some people. Authors will have to decide whether to adjust or just power through that suspicion.