“Hey ChatGPT. Do you like this cover?”

“What’s the best way to launch a book?”

“Should I buy an ISBN?”

“Can you find me an authoritative quote about the importance of self-care?”

Do any of these prompts sound familiar to you? If you’re an author (aspiring or accomplished), you’ve probably started to experiment (even if begrudgingly) with how AI can aid in the publishing, promotion, or writing process. You might just use it for research like you would Google. Or if you’re a little more seasoned, you may be starting to use AI in a more generative fashion: designing cover concepts, critiquing your manuscript, or even assisting you in filling out a scene in your novel. 

Regardless of where you stand on the AI-spectrum, it’s becoming clear that AI is here to stay, and all of us writers need to figure out how to use it to enhance (rather than inhibit) our work. This reality has a lot of people rattled and understandably nervous. In my travels to various writing conferences this fall, I heard plenty of concerns about the role AI plays in our future. As with every new technology, there’s a fear that it will “replace” human ingenuity, creativity, and profitability.

I, for one, am not scared by AI. Just like fire, the printing press, or the smartphone, I see it as a tool. But I am a little scared about the way people are using AI. I can no longer count the number of times I’ve heard people I know and/or work with making dubious assumptions about AI’s abilities. I’m noticing that many writers are often unconsciously ascribing undeserved God-like abilities to their favorite AI platform. “But Gemini says I can,” goes a common refrain. Or, “I don’t need a cover designer. I’ll just use ChatGPT.” 

In that spirit, I wanted to pull together some of the more common AI mistakes–or mistaken assumptions–I’ve noticed writers using lately. This is by no means a comprehensive list; and I’m sure it’s full of its own mistaken assumptions. My hope is that it will give you a little pause–and a few tools of discernment–as you engage AI in the writing, publishing, or promotion of your book. 

Here they are, in no particular order. 

Asking its opinion

Note that I’m using the “it” pronoun for AI quite consciously, because no matter how friendly and human-like its voice or tone, it is most definitely an “it.” That’s why I always think it’s a bad idea to ask AI for an opinion. Language Learning Models like ChatGPT don’t have opinions! Or at least none of their own. They are really good at mimicking them. If you ask it if it likes your book cover, for example, it will most definitely give you an opinion. But the opinion will be its best guess at what an opinion might sound like, based on its ability to process a MASSIVE amount of information really quickly, pulling in searchable content from around the web about what makes a good and bad cover, what to look for, etc. But just because it can summarize data, doesn’t mean it can create a real, useful opinion! It can’t actually “see” your cover and give you any real judgement based on experience, wisdom, or a developed aesthetic sense. 

So please, don’t ask it for an opinion. It will certainly give you one, but what you get won’t be trustworthy. 

Seeking critical feedback

Rather than asking for a general opinion of a piece of work, many people will get more specific. They’ll ask AI to give them constructive criticism or suggestions for improvement. This is certainly a better approach than the former, but you still need to take it with a serious grain of salt. 

We recently designed a cover for a memoir written by one of our clients. It’s a deeply personal story with a lot of layers and symbolism, so it took a lot of back and forth to come up with just the right look and feel. But we hit the mark! Not only did we have a cover that met best practices in the publishing industry; we had something that connected to the essence of the story–one that our client was proud of. 

Then they asked ChatGPT if there was anything we could do to improve it. Do you suppose the AI bot said, “No, this is actually really good. I don’t have anything to add.”? Of course not. Like that friend (we all have one) who can’t help but offer their thoughts and criticisms on virtually everything they encounter, ChatGPT offered some “helpful suggestions.” That’s what it’s supposed to do. It can’t ever say, “No it’s great.” 

Now, I’m not saying that there’s any such thing as a perfectly complete cover (or perfectly complete anything for that matter). I’m not even saying you shouldn’t ask AI for ideas about how to improve the projects you’re working on. It’s just that you have to take these suggestions very lightly. As sophisticated as it can be at times, AI often doesn’t really understand many of the complex and intricate layers that go into writing or any other form of art. And it isn’t taking all of the factors that led to any particular outcome into consideration. 

In this particular case, the improvements ChatGPT came up with for the memoir cover were marginal at best, and in many cases downright wrong. But because our client had put a little too much faith in what AI was suggesting, the response caused them to doubt their instincts. It threw a serious wrench into the publishing process. We finally worked it out, but it was in spite of AI, not thanks to it. 

So when you ask for suggestions or criticisms, it’s important to remember that AI is not an expert, but they will always present as such. 

Anchoring your bias

I’m not a graphic designer. I’m a words guy. So any time I need something visual created, be it a podcast logo, a cover concept, or an illustration for the children’s book I’m working on with my son, I have to rely on a visual artist. Of course, good designers are expensive and their work takes time. So with the emergence of ChatGPT, I’ve started to use it to quickly and cheaply generate visual ideas for any project I’m working on. 

In some ways, this is amazing and super empowering! But I’ve talked to a lot of graphic designers, including my own company’s cover maestro, and they are becoming exhausted by the way their clients are using AI. Why? There’s a little thing called anchoring bias that happens when you use AI to generate ideas. These robotically produced concepts become anchored in your psyche. You then compare everything else you see to these first, exciting graphics you were able to generate seemingly out of thin air. So when your designer offers something different (and probably much better), you can’t view it objectively. 

I’m not saying you shouldn’t use AI to help generate visual ideas. But please be aware of your biases, and do your best to remain open minded as you work with real-world experts in the graphical field!

Assuming it understands your unique situation

One of the biggest blind spots I’ve found with AI is that it’s great at giving you general guidance, but often fails to take into account the important unique details of your situation. 

For example, one of the most common queries around book promotion is whether or not someone should run a pre-order campaign. There’s a ton of content online suggesting that pre-orders are a critical aspect of the book launching process. So when you ask ChatGPT about running a pre-order campaign, it’s likely going to draw upon all this information and suggest you do it. AI’s conclusion will probably be that you should run one because the “general author” might benefit from it. 

But you’re not the general author and it doesn’t know what it doesn’t know. It may not know, for instance, that it’s your first book and that you’re self-publishing, so the “bump” pre-orders give some books probably doesn’t apply to you. It may not take into account that you have limited marketing resources and may not want to waste them on pre-orders due to their dramatically lower ROI. It may only focus on the reasons why to run a pre-orders campaign, and not the reasons you might skip it (here’s an article I wrote on why it’s usually a waste of time for most authors). 

As you engage with AI, it’s important to remember that it can only work with what’s out there–and what’s out there is usually going to be pretty generic. Generic advice is a good starting point, but it often fails the “specificity test.” So make sure you’re testing its hypothesis, using follow up questions with more unique details that might shift the calculus. And don’t forget the experts! People are much better at empathy than AI and will be able to understand the nuances of your unique situation. 

Mistaking its confident tone for certainty

One of the most deceptive aspects of AI is how certain it often comes across in its answers, even if they’re completely wrong. Many of us have had the experience of asking it a question, getting a firm answer, then adding a few more pertinent details, and getting a contradictory answer with an equal amount of certainty. 

Recently, during the research process for a business book I’m working on, I asked ChatGPT to find me a good quote about scaling a business. It found me a perfect option from Verne Harnish’s book, Scaling Up: “Scaling a business is not about doing more things, it’s about doing the right things consistently.” Bingo! Of course, when my editor later fact checked the quote, they found it to be incorrect. The quote, it turns out, was a paraphrase of Harnish taken from a Goodreads review of his book. Thanks a lot AI.

Remember, AI is here to please you and impress you. It can’t say “I think this might be the way to go, but I’m not sure.” It delivers its robotic knowledge with confidence. 

Trusting without verification

The lesson from the scaling quote above (and pretty much all of these points for that matter), is that you can’t fully trust AI. So I’m going to re-phrase the title to this section: Don’t trust and always verify. It’s that simple. Even some of the most passionate AI entrepreneurs who have convinced private equity firms to invest hundreds of millions of dollars in their technology admit that AI isn’t 100% accurate. In fact, it’s even known to “hallucinate,” making up things that aren’t even there. So all of us need to act with a tremendous amount of caution, even as we marvel at what this emerging technology can do. 

It reminds me of the early days of Wikipedia. It became clear pretty quickly that while the crowd-sourced encyclopedia was a great way to do basic research, it couldn’t be counted on for 100% accuracy. Because there was no “authority” reviewing the information, it couldn’t be fully trusted. The rule of thumb in college was that we couldn’t cite wikipedia in any paper, but we could use the list of sources in any given entry. 

It’s similar with AI. In most cases, it’s not creating original material. It’s synthesizing what’s already out there. So while it’s a useful tool for this synthesis, it will always require deeper investigation into the source material. It will always require the extra context that comes from human-based intelligence. I truly believe it can enhance, but never replace our intelligence. Keeping that in mind as we use it is critical to navigating this brave new world.