Hi there,

It’s time to talk about artificial intelligence again! In last week’s Reedsy marketing newsletter, I looked at how likely AI is to impact the writing process. (Spoiler alert: in its current state, I see it as more of a distraction than a helpful writing tool.)

That said, I think AI can be used effectively on the marketing front right now.

If you read last week’s newsletter, you’ll know that generative AI excels at short text generation. In other words, it’s really good at writing copy — and copywriting is a big part of what marketing is all about.

Need to write an Amazon book description? That’s copywriting.

Need to come up with a great tagline for ads? That’s copywriting.

Need to write a newsletter to announce an upcoming launch? Yep, that’s also copywriting.

Ironically, while writing novels (or nonfiction long-form content) and copywriting are both about putting words after one another, they are entirely different skill sets. Which is why most authors hate writing their book blurbs (or any marketing copy in general).

In upcoming newsletters, we’ll see how authors can leverage current AI technologies (in particular, ChatGPT) to help with marketing copy. But first, an important caveat about the use of AI technology, and the underlying ethics of it.

The ethical (and practical) implications of generative AI

There are many legal and ethical questions surrounding AI right now, but most of them stem from a specific issue: how these “machine learning models” have been trained so far.

ChatGPT, Bing’s chatbot, Google’s Bard, and similar chat-based generative AIs have all been trained over many years on a huge corpus of online texts, including:

• Websites and blogs
• Books and academic papers
• News articles
• Social media platforms
• Forums and discussion boards
• Etc.

If you’re curious about which specific websites were used the most for training purposes, The Washington Post wrote an excellent, in-depth article on AI chatbot learning — where you can even check whether your website/blog content was used for AI-learning purposes.

Now, the key thing here is that at no point did any of these websites, news sites, or personal blogs expressly agree to their content being used by OpenAI, Microsoft, or Google for the training of their AIs.

Of course, they didn’t object to it either. It’s just that the question was never really asked, because generative AI wasn’t a big thing until recently.

Interestingly, Scribd.com, a leading subscription service for books, audiobooks, and magazines (to which many of you probably distribute if you use Draft2Digital), comes up third in the ranking of domains used for AI-training purposes.

Source: The Washington Post

Scribd’s terms of use currently expressly prohibit any Scribd user from using “any portion of the content on Scribd for purposes of training a machine learning, large language, or statistical or data model.” But was that language already in place when these models were trained on their content? I don’t know.

More worryingly, b-ok.org, a “notorious market for pirated e-books that has since been seized by the U.S. Justice Department,” is also high on the list, in position 190.

What this all means is that these amazing generative AI models have effectively been trained on content for which they weren’t granted a license — in some cases, where the copyright owner prohibited such use. And in other cases, they were trained directly on pirated works.

I said in last week’s newsletter that I had major ethical quandaries with AI, and this is really the big one. While it is up to the legal regulators to rule on this (which will probably take years), I do believe that any of us talking about using generative AI chatbots have a moral and ethical obligation to touch on this, and raise these questions.

Whether you then choose to use AI chatbots or not is up to you. I personally do (I’m happy to explain why if you’re interested), but I certainly won’t look down on people who decide not to.

In any case, what you should bear in mind if you use generative AIs is that anything you say may be held against… well, no, may be used for further training of the model. If you paste in your whole first chapter into ChatGPT, that content is now in its database. Same for any copyrighted content, or confidential information.

Again, you may or may not be fine with that, but what is important is to be aware of it.

Note: this AI-training issue is of even greater importance (and controversy) for image generation AIs — but more on that in an upcoming newsletter.

Using ChatGPT for advertising copy

After this brief but important ethical digression, let’s look at one immediate and simple use you can make of AI chatbots for your marketing: coming up with advertising copy.

This requires next to no creativity or knowledge of prompting, so it’s a perfect exercise to start playing around with ChatGPT (or Bing, Bard, etc — you can pick your poison).

The idea is simple: you feed the chatbot your book description, and ask for it to come up with relevant advertising copy. You can specify the length you want, as well as where you will use it.

• Step 1: Paste in your book description, and ask the chatbot to familiarize itself with it. You may also paste in editorial reviews, or particularly well-written reviews of the book for further context.
• Step 2: Ask the chatbot to create the advertising copy you’re looking for. Specify what platform the copy is for, as well as any character or word length requirements.
• Step 3 (optional): If there are any results you don’t like, ask the chatbot to replace them. Similarly, if there is a concept you particularly like, ask the chatbot to iterate on it.

Using GPT-4 to generate headlines for Facebook advertising

As you can see, not all the options the chatbot spits out will be great. In fact, most probably won’t. But you can keep the ones you like, and repeat this exercise to get more.

“But why would I want so many different headlines?” Well, for dynamic creatives, naturally! If you don’t know what these are, it means you didn’t take our free course on Facebook Advertising for authors. Fear not, though, it’s still available here.

Instead of the exact copy you’ll end up using, you might sometimes just get a concept from the chatbot, or inspiration for something else. For example, I don’t think “outsmart competition with Amazon ads” is a great headline, but it’s a good concept that I could work into the book description or the primary text of the ad.

All in all, you can see these chatbots as creative assistants who can tirelessly spit out ideas to you. Most won’t be great, but some might ignite the creative spark you were looking for.

That’s just one example of how you can leverage generative AIs for copywriting. Next week, you’ll get some more.

Until then, happy writing, and happy marketing,

Ricardo

Copyright © 2025 Reedsy, All rights reserved.