Writing with A.I.

I know I haven’t blogged much in a while. But don’t worry, I haven’t written much in a while either. Between multiple surgeries in 2025, starting a new job in 2026, and a host of other obstacles, writing has continued to take a backseat. Not necessarily by choice but by necessity. But now I’m trying to get back into writing because I have so many stories I want to finish telling and so many more that I want to start telling.

However, in the time that I’ve been away from the keyboard, technology has undergone its latest revolution. And only those of you who have spent the last few years hiding from technology wouldn’t know that I’m talking about Artificial Intelligence, or A.I.. Services like chatGPT, Claude, Chapter.ai, and others have flooded the space. Companies are integrating A.I. into their products and leveraging A.I.’s productivity boosting capabilities into their day-to-day processes. Odds are that you’ve either used A.I., even if you don’t know you’ve used it, or A.I. has been involved with something you use. People are creating applications with A.I., art, documents, and more.

But the problem that A.I. has created, specifically for authors, is that now non-writer people are using A.I. to create content. And by content I mean novels, poetry, essays, research papers, etc. that they then want to publish. For an author doing vanity of self publishing, this isn’t too big of a deal because neither of these avenues typically care about A.I. generated work. The caveat to that though is that A.I. generated work can include content from other writers since that’s what the A.I. models were trained on. This means that if you use A.I. to create a novel for you and think that you’ve created the perfect book without actually writing it, there’s a chance that the A.I. model you used created that perfect book by using portions of work written by someone else. This could put you and your book in the crosshairs for potential plagiarism claims from the original author.

Now, the bigger problem is that now authors looking to take their work to agents for possible traditional publishing deals are being shut out by agents if any part of their work has been touched by A.I.. Say you spend months, maybe years, tapping out your masterpiece and go to write your query letter or synopsis but opt to let A.I. generate that for you…well, odds are the agent will toss your submission because of it.

It seems that many agents are 100% anti-A.I.. They don’t want any submissions that have been touched by A.I., even if only something as simple as the query letter of synopsis. I assume it’s because if you’re willing to let A.I. do that work then they can’t trust that you didn’t let A.I. do more work in the manuscript. I’m sure a lot of that is concern about plagiarism and whatnot but at the same time, A.I. models tend to all write the same.

Truth be told, A.I. models are trained on existing work. This means that they learned how to write novels by studying published novels written by other authors. This is what the plagiarism concern is so real but also why A.I. generated novels have a distinct “sound” to them. When you read something written by a human and then something written by A.I., the A.I. version tends to follow a very reliable set of patterns and always seems to drift toward the supernatural. You could tell it to create a novel about car chases and it would drift into supernatural mysteries about alternate realities and the car chases just become a backdrop to the story rather than its focus.

And I’m not going to lie, I have tested various A.I. models to see what their writing capabilities were and if they could emulate my writing style. The results were disappointing across the board. I could provide them complete novel storyboards with detailed plot lines, character descriptions, chapter outlines, and examples of my writing to emulate and every model failed every test. The stories created would deviate wildly from the concept provided. The length was usually significantly less than what was specified. The writing style was not mine but clearly A.I. generated. I could spend days, sometimes weeks, trying to tweak and rebuild stories by reinforcing writing models, storylines, etc., but the end result was always not what it needed to be.

Honestly, it would be easier and faster to just write the stories than to attempt to have A.I. do it for you. Besides, there are other tools out there that look for the tell-tale signs of A.I. writing that are almost impossible to beat. Of course, there are ways to “humanize”  A.I. generated content so it is less likely to be caught by standard checks. But don’t forget that those “humanize” services are essentially just A.I. models looking for several of the standard signs that many detectors use and then editing those using A.I. so you’re using A.I. to hide A.I..

To me, that’s like shooting someone with a gun to hide the fact that they were shot with a gun. That’s not to say that it might not work but I have a hard time trusting that if an A.I. can mask its writing so simply that it wouldn’t do that from the start. Why not have the models trained to write in a “humanized” way from the get go? It just seems questionable as to how A.I. knows how to not make it read like it was written by an A.I. but doesn’t use that as a standard. Perhaps that’s by design from the developers but it just seems like something that doesn’t make sense to me.

But the moral of this story is, even though A.I. can be great for helping to summarize and assess the quality of YOUR work, don’t let it do any work on your behalf, especially if you plan to pitch your work to traditional publishing agents.