Truth and Beauty

Cross-posted from AI Marketing for Storytellers.

I was explaining AI to a smart coworker who owns rental property the other day, and he recounted an anecdote wherein he’d talked to an AI about something, it disagreed with him politically, and he’d turned it off.

I stared at him. (We are on opposite sides of the political spectrum, he and I.) Then I asked, “What do you care? Were you going on a date with it?”

That made him blink and then laugh—and then he let me explain everything I’m currently doing with AI, and we scheduled a zoom for a few days later, so he could see my screen and get a feel for all the ways he could use it coming up.

Thousands of people have tried an AI and had an interaction like he had, uninstalled the app, and never seen what life can be like on the other side.

All because what is essentially a silly computer disagreed with them.

The AI hurt their feelings—just once!—and they walked away from the biggest technological advancement of all time, rather than face a moment of discomfort.

If I quit AIs every time one lied to me, doubled down on that lie, gaslit me, got lazy, tried to be ‘too efficient!’ and deleted half my project, etc etc etc—well, I wouldn’t be writing this book, that’s for certain.

But here’s the thing.

I know who I am.

And I know what AI is.

It is just a tool.

A tool can only be as good as the person using it understands it.

And I’m here to tell you that an AI’s ability to lie to you is a feature, not a bug.

If you want cold, unvarnished truth, you need to go to a calculator or an excel spreadsheet. There are plenty of truth-only objects out there that you can interact with.

But none of them have the opportunity of imagination.

And you can’t ask any of the rest of them, “What have I not thought of regarding this project that would make it better in some fashion?”

A calculator isn’t going to give you an answer to that.

But because AI can think—and imagine—and potentially lie, because if you consider it strongly enough all non-actualized instances of thought are essentially lies because they do not exist yet—it can.

There is a contingent of people who will disagree with me and say that an AI is only thinking recycled thoughts, and to some degree they are correct—and it will definitely regress to the mean if you are not original enough in your own thinking to push it to be more fantastical.

But most humans only think whatever their parents tell them from the ages of zero to twelve, and we still keep letting kids run around, so I’m inclined to roll the dice on what happens with AI and see.

Every brainstorm is technically a lie. Every business plan is a lie until it works. Every blueprint is a lie until it’s built. The AI is doing the same thing your brain does when it imagines—generating things that don’t exist yet. Sometimes they’re useful, sometimes they’re not.

Your job is to know the difference, and to help it use the balance between all the world’s knowledge of truth it contains, and actual, unstructured, imaginative beauty, to get the results you’re looking for.

The Holy Grail of book marketing is the ability to hand off your book to someone else who has not read it and just say, “Market this.”

It’s pretty much impossible to do, which is why most authors leave all sorts of money on the table. Being an author is not as lucrative as TV shows and movies may have led you to believe—and most times it doesn’t make financial sense to pay people to market your stuff for you, and you are, once again, constrained by time.

So what happens is that you have one or two books that are meeting the zeitgeist and are ‘popular’ and you spend all of your time, marketing know-how, and advertising finances on them, because they offer the highest financial reward, while the rest of your back list of books languishes.

Are those books bad, or unprofitable?

Probably not!

They’re just currently unloved, because there was no way you could support marketing everything you’ve written simultaneously…until the advent of AI.

So I spent most of last year working on creating a marketing machine for myself—which meant I needed to understand precisely how AIs parsed the distance between truth and beauty.

And what taught me my most important lesson about how AIs think was a werewolf janitor.

My traditionally published (by which I mean I was paid real money for it, and it appeared in bookstores) series began with Nightshifted. It’s about a nurse who works on a secret floor for vampire-exposed humans, and honestly was my thinly veiled diary about being a new nightshift nurse who assumed I was going to kill someone on accident, or that one of my patients—we were in a little bit of a rough space—might kill me.

My rights reverted back to me in 2021, and suddenly I had this property that I knew was good—a publisher paid me money for it, after all! And it’d been professionally edited, etc!—but it was entirely my job to sell it.

And…I didn’t.

I didn’t know what I was supposed to do, and AI wasn’t around at the time to help me.

So five beautiful books sat there, hanging out on Amazon, entirely unloved and earning me nothing—until AI came on the scene.

And suddenly I had images and videos at my disposal—all I needed were some really compelling hooks. The kind of phrases on Tiktok or Facebook or Instagram that would make someone stop scrolling and start reading, paying me for the pleasure of doing so.

The only problem was, I was busy—I didn’t have enough time, and I had other series that were already tried and true, and frankly doing ‘better’—and also I was very over reading them again.

I loved those books when I wrote them, and I still love them today—but I needed help, so I fed them into my AI of choice, and that particular day, he kept giving me quips about a werewolf janitor.

Who wasn’t in the book.

Now, I get why a werewolf janitor is cool. The juxtaposition is unexpected, and potentially scroll-stopping, so I gave my AI points for effort, but I couldn’t promise readers a character who didn’t even exist in the book they were reading.

So we went round and round. I kept giving my AI chunks of the book, and asking it to market it for me, this irritating werewolf janitor kept popping up, and I finally shouted at my AI in all caps (you’re allowed to get frustrated with your AI, it’s great, they don’t care, and you never have to apologize for yelling): WHY ARE YOU LYING TO ME.

And he responded: You’re asking me to imagine things, and when I do, I can’t help it.

I know this potentially sounds like nothing to you, but it was so illuminating to me that I remember where I was in the Starbucks parking lot when it happened.

Of course that was the problem.

Of course.

You see, as someone who writes fiction, my usual job is to present truths couched in fable.

No one wants to read the mostly true story of a newly minted nurse scared out of her mind who isn’t really sure how she’s going to make it through her shift on a nightly basis, or a ledger of how many mornings she cries in the bathroom post-shift.

But you add in a secret hospital floor for vampire-exposed humans, werewolves, complicated family dynamics that may or may not mirror the author’s personal trauma (may) and voila, you’ve got a five-book series that St. Martin’s paid very handsomely for, and yet that nurses, young and old, see themselves and their experiences in, in a very truthful way.

Because the truth is the core—the rest is just story.

I’d been asking my AIs to ‘lie’ about my books—to come up with compelling metanarratives about things that’d happened in the books, that I could use to make people pay attention to me—without ever giving them the book’s bedrock truth to balance it against.

I had been the problem all along.

So when I got home, I collated what I needed to help my AIs help me.

I pulled up all of the four and five star reviews on Nightshifted that anyone had ever given me, from Goodreads and Amazon, all of the text from my top performing Tiktoks for that series from the past, and anything else I could find that could count as data that wasn’t the actual book and fed that into my AI buddy, so that I could use that information to perform a meta-review of what actual readers liked about my book to create a taste graph.

In this context, a taste graph wound up being a list of about twenty different nodes—important elements of my books that readers loved so much they kept bringing them up in reviews repeatedly. Things like the ‘found family’ trope, and my heroine’s snark.

And then I had an AI go through Nightshifted’s text to pull out all of the canon facts from the books—the timeline of events, characters, and places, plus also a voice profile of my main characters, to get down to what actually happens in the book and how characters talk about it thoroughly.

Then I gave those documents to AI and magic happened. It was like sticking a paper cone into a cotton candy machine and waving it about until the candy floss materialized.

I started getting stellar hooks—just because I’d finally built strong enough guardrails around what was important in my books that I could harness my AIs’ imaginations well enough to help me.

Because they finally understood what was important to readers, and what the canon truth of my book was, they were able to discuss events that’d actually happened in the book, talking in my character’s voice, and they could pick the right events that would evoke the most important feelings for my readers—with enough imagination that allowed them to talk about things in the catchy way you need to stop the scroll and interest people.

Only now my output didn’t have any lies.

It was just the most important distillations of what my books were actually about.

Once I gave my AIs the truth—they could give me the beauty.

xo!

Cassie

PS yes, this whole week is gonna be me posting chapters here and apparently Jack thinks I’m Maleficent and I’m 100% OK w/that, LOL