Disclosure: I may earn affiliate revenue or commissions if you purchase products from links on my website.
Promotion: I recently published Framed: A Villain's Perspective on Social Media. It's an outgrowth of everything visitors have liked about this blog since 2016. It's now on sale for $0.99!
Two weeks ago, I published Framed: A Villain’s Perspective on Social Media. It’s a provocative book and I was prepared for the full spectrum of feedback. Then people started asking me about AI. They wanted to know what role it played in the book’s creation. This wasn’t something I was prepared for, even after writing a book that contained a lengthy chapter about the “dead internet theory.”
Is my book so good that people assume it isn’t an original human work?
Is the writing so formulaic and redundant that it seems generated by a robot?
Are there unnatural design elements in my cover?
Are people just fishing for tips on writing and self-publishing?
AI small talk is about as common as commenting on the weather, but people might not realize that bluntly asking about AI’s use for a creative work may be considered offensive. The connotation that I used generative AI to write Framed isn’t far from me having plagiarized or worked with a ghostwriter.
Framed is an original work and generative AI was not used in its creation, including for “workshopping.” I was a bit paranoid about this, so I ran large chunks of text through plagiarism detectors. I can stand on business here. There is plenty of my writing out there on the internet. If it’s compiled and compared to Framed, the style and idiosyncrasies will match up.
On one hand, AI writing tools can help level the playing field between self-published and traditionally-published authors. However, the pervasiveness of AI is starting to degrade what it means to be an author. Rather than going into the conspiracy theories and social impacts as I did in “The Puppeteer Part II,” I’ll dive into my thoughts on AI as a first-time author who started writing material before 2020 and who published his work in early 2025.
Most writing tools are “AI”
The first time I was asked if I used AI to write, I snapped, “Well, I guess I was using AI in 2012 when I purchased a Grammarly subscription!”
Grammarly now prominently features “AI” in its branding. While artificial intelligence has come to encompass so, so many things, Grammarly certainly uses LLMs, so it’s accurate to classify its products as AI. I guess I was using AI in 2012 when I earned a B in English 001.
My classmates and I saw Grammarly as a slight upgrade to Microsoft Word. Today, there are so many tools and aids and automations that there exists a class of AI-enabled “writers” who couldn’t produce anything near the same quality of work if using a pen and paper. Written content creation now benefits from tools well beyond spell check or grammar suggestions.
My tech book was low tech
I wrote Framed entirely using Google Docs. It’s a bare bones digital writing experience–there isn’t even a “dark mode.” Docs’ features resemble what I grew up using: spell check and grammar/usage suggestions.
I like Docs because I was able to pull my phone out on the train and quickly pick up where I left off. There’s also something to be said about the app being browser-based. There are more distractions, but shifting between so many Chrome tabs allowed me to quickly (and, intuitively) synthesize information.
By early in 2024, I had made some big mistakes where I had either returned to the wrong “branch” of a chapter or had lost important sources. I needed spreadsheets to organize dimensional data. My big upgrade was transcribing my sloppy checklists into a massive Google Sheet. Today, that sheet has more than one thousand rows covering potential sources and works cited, which were compiled by hand.
The funny thing is that I was developing natural language processing (NLP) programs and tools before generative AI was a thing. I largely avoided using AI or any advanced software assistance for this project.
However, there are a few areas where I dabbled, and I feel that they’re worth talking about.
Copy Editing
The manuscript of Framed exceeded 150k words. Before my editor touched anything, the average chapter length exceeded 4000 words. The word limit for ChatGPT is 4000 words, and I only know this because I tried to paste a few of my gigantic chapters into ChatGPT to get some feedback.
When I was waiting for my editor to send her work back to me, I tried using AI as a copy editor and as a developmental editor. Though it certainly identified issues, especially those related to awkward phrasing and choppiness, it wasn’t anywhere near as precise or consistent as my editor.
I also found that it often jumped to rewriting huge swaths of text without being prompted. This behavior dangerous for the developing writer, as it encourages lazy habits and, ultimately, may result in “singularity” of writing style.
Ultimately, I don’t fault anyone who uses ChatGPT for light proofing or copy editing. For me, it wasn’t anywhere near as thorough as my editor, and it made a lot of annoying mistakes.
ChatGPT is a godsend for immediate feedback…but…
The biggest issue I had with Framed was getting developmental feedback. As in, “Was this chapter…good?”
When authors talk about beta readers and early reviewers, they don’t actually want a multi-week feedback loop. They want feedback overnight! I was so desperate for this feedback that I began offering friends money to read chapters and to “roast” me. Those offers mostly went unfulfilled.
I took to asking ChatGPT. Yeah, it was useful and it helped me stay grounded. However, it was way too positive. Everything I wrote was “amazing,” my “wit” was “excellent,” and so on. I asked it to be more cynical (and realistic). Past that first scan, which did generally tend to be useful, we were going in circles.
I think I ran it for four of the shorter chapters. I found that the feedback centered around stylistic peculiarities or aspects of my voice or writing style, things that, if changed, would have decreased my own closeness to the story.
Ultimately, I don’t think it’s dishonest for a writer to use ChatGPT to give developmental feedback on an original draft. I think the problems creep up when ChatGPT is generating the ideas, ChatGPT is generating swaths of the text, and then ChatGPT is editing what is, essentially, ChatGPT’s work. The term “overfitting” may apply.
Blurbs, promotional material, ads, and pitches
AI is being used heavily in the marketing space, if that wasn’t clear.
I paid one consultant, just for him to send me sample blurbs which were just AI-spun snippets from my existing material. Not cool man.
There are all types of services borne out of this AI rush. There is one thing I agree with: humans, even humans who are excellent writers, aren’t necessarily great at writing marketing copy. It’s a skill that must be developed. Unfortunately, with this skill, “learning” and getting feedback usually means losing lots of money in ad cost and missed opportunity cost while optimizing the writing.
I’ve run ads on ad platforms where AI-generated ad copy is all-but-forced on me. It’s a weird state of affairs.
I wrote all my blurbs, headlines, ads, and even my pitches. I’d like to think it helps me stand out in a sea of generated AI, where AI stands for Absolute Idiots.
Cover Design
I didn’t use AI for any aspect of cover design. At first, I wanted to hire a designer. I created some wireframes in Photoshop and then contacted a few designers. They either ignored me or produced work and suggestions that I didn’t jive with.
I went back into Photoshop with the thought, “No…I can do this.” Social Media isn’t a hard genre to design book covers for. I tried using AI for some illustrative elements, particularly the original heart motif, but it looked awful.
Dissecting my cover:
- The heart icons were drawn by me
- The “Create post” button graphic was drawn by me
- The fonts, I don’t own, but are free for commercial use
- Everything else is pretty much just basic shapes, Photoshop features, color theory, and some artistic liberties when it comes to negative space and such
I’m strongly opposed to AI being used in cover design. The lesser ranks in my genres are filled with AI crap. However, I understand that there are some genres in which the “standard” illustrations aren’t things that the average writer with Photoshop can draw up.
Research & Citations
One thing emerging in the indie author communities that I follow on Facebook is that people like using “AI” for research because it tends to produce different results than Google.
Though I pretty much only used AI for research to “restore” old sources that I had referenced and then lost (especially in cases where I had referenced archived pages), I found this to be true. Google results are spoiled by the SEO game (addressed in Chapter 4, “Algorithms and Truth”) and “AI results” were sometimes both extremely relevant and containing things I had never seen before!
However, sometimes ChatGPT cited “facts” that challenged my own research or experience. When I asked for a source, there was no source. It’s nearly impossible to trace the logic that could lead to, essentially, rumors or low-credibility discussion, being upcycled into facts.
There’s a certain point at which using ChatGPT for research undermines what it makes to research. However, there are probably researchers out there who would scoff at my use of google filters and the Internet Archive. I take pride in the immense amount of research I did before realizing that ChatGPT might have been better at stitching it all together than I was. I barely used it, but, after using it, I’m less judgmental of writers who do. My larger concern is that it can easily then lead researchers to bridge ideas that they wouldn’t have otherwise considered. That does degrade the creativity involved with writing.
I used ChatGPT as an interactive formatter for formatting Chicago-style citations for my work. To be honest, it made a ton of mistakes that I had to correct, and I should have instead used a built-for-purpose citation builder.
Takeaways and the future of writing
Using AI to help with a self-published book can mean so many different things. If most people could view a replay of how I constructed Framed, they’d probably conclude that I underutilized the “acceptable” AI writing tools. Maybe my book is worse as a result, but it sure as hell is original.
In reviewing the work of other first-time authors, my impression is that everyone is overusing it. Covers, blurbs, intro chapters–I’ve even seen some subtitles that I suspect were AI generated. Because AI is going to become better and better, it’s also going to become harder to detect.
This makes me sad, but for the aspiring author, it should be a sign to start your book now, before it really is too late. If you’re overburdened by thoughts of how AI will destroy humanity, pick up a copy of Framed: A Villain’s Perspective on Social Media. People are saying that it’s a neat distraction.
