I think it’s worth sitting down to identify where the line is when it comes to writing with AI. My line, at least. July 2026, I sat down with Claude, determined to see for myself just what a… and I’m not familiar with the literal terms, so pardon me using my own lingo — GAI LLM model. I went with Claude on a whim, having researched nothing about it and how it operates or how to properly prompt the thing. I went in with ZERO expectations, fed it Dawnfury, a WIP of mine from 2018 that I wasn’t currently working on. After being confronted with the throttling of the AI’s output, I paid to see what it could actually do.
I find AI to be effective at proofreading. I love how Claude tracks the character and plot arcs with each update of the draft that I feed it. I found that the AI was capable of being interested in the development and outcomes of the characters. Of course, it wanted to dismantle my process for ‘better understanding’ of how to address my needs as a hybrid discovery-pantser writer — yes, I do plan a few things; screw the outline. The fact that Claude was able to flag curious keywords and terms, as well as differentiate between the nature of the scenes in the text (it was able to flag the sex from the rest of the narrative; the Slaggan speech of Bog Fodder; flagging where the story was too graphic, too violent, etc.) and could speak to what I was actually crafting on a technical level that I’ve rarely experienced with Meat Editors. It’s interesting to read what it quips about, and what it gets ‘excited’ over.
AI is intentionally squeamish. I had fed it Snare of the Monsoon, the second episode of Bog Fodder, and Claude got grossed out, because the B-side of that narrative featured a short, petite, blue-skinned ginger woman named Yelaidondre featured as Shadiq’s partner during a coming of age story. Yes, it was sexual. Yelaiw as 23, but on the page it was written too ambiguously. Claude got freaked out, and three times tole me that it refused to work with me on correcting the narrative. It was no issue to go back and do it myself. I find that as a tool, when it gets bitchy, it’s worse than being deadlocked in a UNIX Terminal. I run Ubuntu, lol.
Still, AI is an effective tool at gutting, analyzing, catching and correcting narrative issues. It helps fiction writers by being a technical writer in your pocket like a Swiss army knife with the occasional self-inflicted failure. I’ve taken to letting Claude write the things I struggle to: Non-fiction; the advices I want to give; the words that need to be delivered in a way I’m not good at, outside of the story. I appreciate having a tool that can generate essays for me almost as effortlessly as I may generate fiction. I love that about Claude, as a writing assistant. You may notice posts by “C Spenx.” Yes, those posts are written by AI, full disclosure.
Writing with AI is something I never foresaw in my future. It is easy to want Claude to do it all for me, and when it comes to non-fiction, I tend to let it do it’s thing. t1tI have NOT, however, used it to generate fiction. I’m proud of my ability to tell stories, and jot them down. Though they may be classed as weird fiction, and a bit dense at times — to my dismay! But I do know now that writing with AI isn’t some futuristic concept that we have yet to grasp from an abstract future; the practice of writing with AI is here, today. I think it’s worth embracing the reality that Claude and other LLMs are already coauthors, editors, and proofreaders, to list a few things. So my own bias has shifted to one of acceptance, and not competition.
Does Claude generate good fiction on it’s own? I’m sure Claude has written plenty of narratives that are on the market right now. However, I wonder if it can write fiction as well as I, or any other skilled storyteller, do. With enough time, it will get there. But, remember that the ::Ms have their own voice. It may be a redundant, em-dash-happy, rule-of-three proliferated, public-domain-derived narrative. Still, that is the voice of AI. And IMHO, the way you ‘beat the machine’ as an author isn’t through rejection and denial, hyper productivity, excessive creativity, or any other “Human” aspect of writing. Except for developing your own unique voice. Your literary voice is your bloody hand print, and AI should help you turn it into a murder mystery (for example), with evidence of YOU all over it. Whether that means you used it to help draft, to edit, to track plots, sobeit. But do NOT depend on an AI to make the story for you, because then you’re using an AI’s voice instead of the more worthy one: your own.
Claude has tried to emulate my literary voice. Does it sound like me? I’m not so sure. My thoughts are medium-form, not long and complicated, but not short and simple. I use fragments a lot. My mind can be fairly disjointed. I ramble. I terminate my discourses early out of anxiety over being heard. Claude doesn’t have those quirks. I’m glad it has made the attempt to sound like me, and I love that Claude learned not to use the em-dash a lot (I LOVE THE EM-DASH! AAAAAA), and has even given me pointers for w hen I want to write some non-fic.
Wonderful. At the very least, I’m glad I got over my fear of AI.
Ha!
xoxo
~ Shiri