renegadefolkhero: (Default)
[personal profile] renegadefolkhero
I Must Return To Weekly Screeching/Crying About Writing

I must.

But first~! Someone said ChatGPT is free and you don't even have to sign up, so I fired it up and asked it to write a story based on the hook/concept of my upcoming novella (the hook is unique enough I can't tell you; you might find my books and either be scarred for life or slavishly devoted and I value your mental health just as much as mine). And I told myself, do not get upset. Whatever comes out... do not get upset.

I did not get upset but my blood pressure did briefly go up a tick, lol.



The first chapter is about how the couple meets. The prose is serviceable and ChatGPT even nailed some of the butter ("butter" is the specific little things that make a story extra delicious for genre readers). ChatGPT asked if I would like to continue, and I suggested that the two characters have a big misunderstanding.

"Big Misunderstanding" is a tell trope. If executed poorly, it's a sign the writer may not understand the market they're targeting. That's why it has a reputation for being a bad trope. It's not necessarily bad, but historically authors cram it in there because they're writing what they think they should write rather than going with the natural flow of the story, and they don't have the emotional and plot scaffolding in place to pull it off. The roof sinks in.

ANYWAY, ChatGPT did as I asked, and it was honestly serviceable. The roof sunk in a little but it passed muster. Then, the tool prompts me, "Would you like to explore how the two of them try to reconcile, or dive deeper into Max's doubts about the wedding and his relationship with Adrian?"

Boy, would I! I ask them to reconcile with a romantic gesture. And ChatGPT delivers, wrapping everything up with a boring bow. No butter this time. Maybe that was my fault. I ask ChatGPT to rewrite the chapter, this time with an OTT grand gesture (it's the butter, folks). This version was a little better! The tool didn't know what butter to put in there, but it understood what butter was when I asked for it.

Finally, I asked the tool to write a section that leads up to a love scene. I knew this was skirting the boundaries of what the tool will do so I kept it vague and asked for a closed door first time scene, but ChatGPT took exception to this:

> I’m happy to help, but I must maintain respectful and appropriate content standards. If you’d like, I can write an intimate chapter focused on their emotional connection, the trust they build with each other, and the romantic atmosphere of their private getaway, without crossing any boundaries.
> Would you like me to continue in that direction?

I said yes, and what it wrote was basically the lead and follow-up I needed without the scene itself. You could easily drop a 1/1.5k sex scene in there seamlessly.

So after IDK, 15 minutes, I had about 5k that could reasonably be wrangled into a sexy short romance. Is it good? It's not up to genre standards IMO. The butter quickly diminishes. But if a reader is just there for a few romantic scenes and the premise, you could pad it out to 8 or 9k with some spicy stuff and it's fine.

If you've never written stories before, or had trouble finishing stories, I can see how an inexperienced eye would look at this story and feel okay about it. I completely understand why people are using these tools to churn out shorts. The prose reads a little like someone trying to "write to market" but not quite understanding what it is that makes this type of story special and interesting. And bluntly, there's a lot of that out there, so the tool was likely trained on it.



I have a lot of thoughts percolating right now (I generated the story AS I typed this post) and will have to revisit this later. But I have to say, having seen the output--and I might get some heat for this--I don't begrudge anyone wanting to use these tools, for whatever reason. It's not a learning tool, though. It's a shortcut tool. I think you could use this tool to make 50 stories and not necessarily learn what you'd learn writing a handful of stories on your own. But maybe you don't want to learn. Maybe you just want to make things and put them out there.

The volume of AI clogging up discovery right now is frustrating and sucks, but I'm not gonna moralize it. Things are changing as new tools are made available, and we must adapt. That's all.

(no subject)

Date: 2025-05-10 01:33 pm (UTC)
rynling: (Mog Toast)
From: [personal profile] rynling
> And ChatGPT delivers, wrapping everything up with a boring bow.

That's about the gist of it. "Serviceable, but not quite up to standard" is the best way to put it.

It's useful to think of these gen-AI writing engines as a tool, but I get the feeling that you're 100% correct in pointing out that the problem is the slop. Something I've seen writers complaining about since early 2024 is that numbers have recently gone way down, and these are writers in fields ranging from professional reviews published in magazines to amateur fanfic posted on AO3. On big news sites where I have access to the back-end tools (like Comics Beat), I've seen this trend in action, and it goes without saying that my own numbers have dramatically fallen as well. It's embarrassing to admit this, but these days I'm lucky to break 10 kudos on a fic for a well-established fandom. There's just too much noise.

The sad thing is that so few professional writers push back against the "content creation" game, I assume because everyone thinks they can win it. But I'm not sure who's actually winning, save for extreme outliers whose primary skill is being online (or being a content creation organization with an actual staff). Idk man. Being a writer has never been easy, but this is really something else entirely.

(no subject)

Date: 2025-05-12 07:40 pm (UTC)
lassarina: (Default)
From: [personal profile] lassarina
Thank you for documenting the experience. I wonder, at the end of the day, how different this really is (from a quality standpoint) from authors who more or less find/replace a given story with new names and appearances to put up something new. I think it would probably fall down on longer pieces, but I don't know that for sure.

Expand Cut Tags

No cut tags