My Changed Opinion on Writing with AI

I’ve spent a lot of time writing this past year. Writing for my job, but also writing short fictions, a novel, and journalling.

For awhile last year, I was heavily using Claude (the AI assistant from Anthropic) to assist in writing. AI is a big part of my job, so I’m inherently curious about it. I was was one of the first 0.1% of ChatGPT users.

It’s a weird spot to be in for me. I’ve always been half technophile and half luddite. Half of me has always been interested in the cutting edge technology and I’ve been an early adopter of many of them, even when I was a kid. But I’ve also grow more and more wary of technology as I’ve gotten older. Probably largely influenced by the folk-punk kids I spent so much time with in my 20’s– but also just because I can see the damage it is doing to our society and environment.

I think technology has a lot of benefits too, though. So it is truly, genuinely conflicting. And that’s the nature of human experience: being able to hold those conflicting viewpoints inside of you.

With all of that said: I’m faced with this moral conundrum… how to use AI in writing (if at all)?

The Way I WAS Using AI In Writing

For the better part of 2024 and 2025, I used Claude and sometimes ChatGPT extensively to help with writing.

Almost never in the way people think, where you just type in prompt and then copy & paste what it gives you back.

I learned that you can get some pretty engaging writing from it if you give it a lot of unique context. Often I would voice memo a bunch of context and then have a specific version of the AI that I had trained to TRY and replicate my voice turn that into an output.

Or sometimes I’d speed-write a draft and then have AI clean it up.

Once I even had it write a whole 60k word novel. I created a project in Claude, then worked through the plot, the characters, the themes, the tropes, etc. Then I said, “just go write it.” I had to give it a few rounds of revisions, but what it came up with after just a few hours was genuinely a good read. Not literary genius, but a fun, engaging read.

I can see how for some genres where the readers goal is escapism and entertainment, this will work very well.

You will see in some of my other blog posts, I’ve added a disclaimer that those posts were editing using Claude. It’s funny in hindsight I can see Claude’s writing patterns all over them, but at the time I didn’t realize how much it was changing from the work I gave it.

I still think my writing is technically better with AI, but I don’t think it sounds like me. And something else doesn’t sit with me…

The Ethical Issues

There are so many ethical issues with AI. For my job, I basically just have to accept these and move on. But for my personal writing, the issues become both more of an issue AND they sit worse with me.

The training data for these systems widely “stole” the work of creators without their permission. It often recreates works that are far too similar to those original creators.

What is really telling is that the major AI companies are creating licensing deals with big publishers and papers. That means they, on some level, recognize that they SHOULD be paying for this content. However, all the independent writers and backlog of training data goes unpaid.

If this changes, where AI companies either choose to or are forced to pay royalties to the creators whose data they train on, then I would be much more likely to use AI– at least in small ways– in my own personal writing.

Then you have the environmental issues. It’s something I care about a lot, but all technology has had a pretty negative impact on the environment and required step-changes in energy use vs. previous stages of technology. I follow climate tech pretty closely, though, and I’m fairly optimistic that there are some profound energy solutions that are on the cusp that AI energy demand is pushing forward. So I’m hopeful. Not naive, but hopeful.

Where I Do (& Don’t) Use AI Now

For one, I believe in being transparent when AI is used. So if I use AI in any of my writing, I will disclose that.

I still use Claude to help as an assistant while writing, especially for my novel. It’s not allowed to write any of the actual prose (or words on the page), but it can still be helpful in pointing out plot holes or inconsistencies. It has been helpful to be able to say “give me a list of all the characters from this group I’ve introduced” essentially acting as a second brain for the project.

Some people might still have a problem with that kind of AI usage, but I think it’s substantially different.

Even when using it as an “editor” I never have it write the words. I will ask it to point out places that need work, and then I’ll go and make the changes.

You’ll probably notice that any new posts from me won’t be as grammatically perfect, but I’m okay with that. This is a blog. I don’t take that much time to edit. I never have in the past, I don’t see why I need to now. It’s more about sharing ideas than perfect prose.

In my longer-form fiction writing, I will be using AI the way I mentioned here and also a mix of human readers and editors. Those take a lot more time and I give them multiple passes before publishing (after an embarrassing publishing on this short story on my Amazon page from years ago that multiple people have pointed out some glaring typos and grammar issues. The story, I think, is still decent, but the writing is embarrassing in hindsight).

Thanks for reading my very human post.