ChatGPT is a Terrible Editor for Your Writing

Why AI can't replace a real editors, readers, and writing friends

A hallmark of TYWI’s Writing Programs is getting continual feedback from their peers. As I lead my workshop of five through 6 weeks of workshopping each other’s work, I remembered exactly why it’s not only important for writers to get feedback, but from the proper channels.

During my thesis research, I experimented with the ways writers collaborated with artificial intelligence as a co-creator, and specifically identified major gaps that made Generative AI tools subpar creative assistants. I myself tested how AI performs as an editor of some of my work. Comparing that with the feedback received from real humans, one thing is clear: ChatGPT is a terrible editor for your writing.

1. AI will tell you what you want to hear, not what you need to hear.

2. AI can catch mistakes, but struggles to recognize missed opportunities.

3. When it gives line-edits, AI falls into non-specific styles, characterized by em-dashes, short sentences that make nifty one liners, and repetitive endings.

4. AI doesn’t have a background, experiences, or feelings. It can identify objective fact, but is unable to empathize deeply with human emotions.

AI will tell you what you want to hear, not what you need to hear.

Generative AI models are designed to be assistants, but my experience, they have been eager, overly saccharine, and hesitant to give harsh, but real feedback. It can create a false illusion that your writing is nearly flawless and publication ready. However, when you get the piece in front of readers, the reality is completely different. This dissonance can help writers feel like they’re making progress, but risks eroding their own ability to critically examine and improve their work. It creates a continual dependency that’s readily available, almost like a dopamine hit. Sure, it can keep you writing, but does it encourage you to improve? Does it highlight where your writing actually needs work, or does it just shroud the truth in inflated praises?

When I had my first couple of chapters workshopped by my TYWI group, there were so many structural, developmental, and characterization issues that AI never even brought up in my trials. Instead, it just said my work was “revolutionary”, “a manifesto”, and “publish-ready quality”. In truth, it’s none of those things, not yet.

Even if AI does give some kind of feedback, you may feel less compelled to take it seriously. After all, it’s just AI, what does it know about anything related to art? See the problem? Even if AI was a good editor, it will never hold the validity of a great human editor, especially not subjective fields. The end result is that you get a lot of useless feedback that doesn’t help you develop as a writer. That’s not a cycle any aspiring author wants to get stuck in, especially this early in their career.

Finally, AI offers the ability for constant, immediate feedback, while human editors take time and often need the entire finished piece. Slow and thorough is good when it comes to editing, and you don’t want to get in a cycle of getting too much feedback, too early. That’s how you get stuck rewriting the first couple of chapters without every getting forward. When you draft, you should do it all in a go. Get the words out on the page. AI makes it easy to have feedback checkpoints at every chapter, and that can be harmful to a writer’s natural creative process.

Human editors and readers, even if you strongly disagree with them, have a “point behind a point” (thanks Rhea from my workshop). Even if you disagree with their solution, the fact that they brought up an issue in your story means something is there for you to look deeper into. They’re often more critical, honest, and constructive. Their criticism may hurt at first. It may undermine your story. But that fire is a good thing when you’re in the editing stage! That collaboration is where the writer-editor relationship really flourishes.

AI needs to be prompted, but often, you won’t know what to prompt.

If we knew exactly what parts of our writing needed feedback, we wouldn’t need editors. We could just identify the problems and workshop them ourselves. But as someone so close to the work, we don’t have that distanced perspective. This means we lack the insight to know what needs help, and thus fall short of prompting. Of course, you could master prompts, but even then, AI won’t really pull out feedback unless you ask for it. Human editors view your writing as a blank slate with their experience and perspectives. They can help identify things you don’t even recognize as problems.

AI is strong at pattern recognition. It can also be strong at identifying issues. In my workshop, I noticed an under-appreciated aspect of human feedback — collaborative creativity. We not only pointed out problems, but understood the story, characters, and intentions well enough to offer appropriate suggestions that spark inspiration within the writer. Beyond this, a human can offer something totally invaluable — identifying missed opportunities.

Sure, you can prompt the AI to give you this kind of feedback. But the AI doesn’t have the same natural creativity stemmed from real life experiences like a human. Its suggestions may be objective, but doesn’t have the collaborative creativity that happens in conversation with another creative person who has an interpretation of your work. I think there’s a lot of magic that happens here, which is why having editors and pre-readers are so important for fleshing out the later drafts of your story.

When it gives line-edits, AI falls into non-specific styles, characterized by em-dashes, short sentences that make nifty one liners, and repetitive endings.

Current iterations of AI have very clear indicators that we’ve heard of. One is the excess use of em-dashes. In my trials with AI, nearly half of the line edits included an em-dash. Don’t get me wrong — I love a good em-dash, but it’s not overpowering my writing. Em-dashes are effective because they emphasize a point, but AI is focused on efficiency, so it seems like it wants to emphasize every point, while using a convention that most advanced writers use.

In my experience, its line edits are a total joke, completely deform my voice, and seems like a caricature of itself. It truly tries to push my work towards AI slop, and no amount of prompting helps it really understand my voice. At the end, it can feel like you’re fighting against it, and that’s just not a productive use of your time as a writer.

Finally, I don’t think it really understands nuance or mystery. If something in my story is hidden or abstracted, ChatGPT can’t really understand and keeps trying to get me to add in that detail. It just doesn’t get it. Human editors also may not get it, but at least they have their own thoughts and feelings that justifies it. When an in inanimate object and algorithm doesn’t get it, it just feels like a waste of time.

AI doesn’t have a background, experiences, or feelings. It can identify objective fact, but is unable to empathize deeply with human emotions.

AI is not processing your story. It’s not getting emotionally invested. It’s not empathizing with your characters. It’s ingesting large quantities of text, comparing them against other works and writing conventions, and spiting out feedback in seconds. How in depth and authentic could this really be when it comes to the things that really matter in a story, like emotional resonance, feeling, and connection to the human experience.

What happens when you rely on AI feedback? Your writing begins to lose your voice, feel mechanic, and become what is known as “AI Slop”, which is just a hodgepodge of valid writing conventions jammed together without thought and consideration. At first glance, it seems fine, good, even. But there’s something uncanny about how perfect and stagnant it is, and that uncanniness isolates the reader and yourself.

Finding Human Writers in the Age of AI

Ultimately, getting feedback from AI is a hollow effort, especially when compared to workshopping with humans. But I’ll admit one thing — it’s hard to get people to read your story. Finding editors is expensive, and organizing beta readers feels impossible, especially when you’re young. When we’re so malleable, insecure, and desire feedback, AI can seem like the quick fix that helps us reach our writing goals.

For non-creative, poor writers, AI collaboration may help them improve their general writing skills and identify glaring issues. But if you’re a dedicated writer working on a longer project, you’re likely better than AI, even if you’re not that good. My first piece of advice is to focus on drafting first. You don’t need feedback after every chapter in your first draft (I say this as someone who really desires this). The first draft is meant to be sloppy and a frantic effort to get all the ideas on to the page. The second draft is meant for you to go through and identify problems you see. The third is to share it with beta readers, and fourth is editors. At least that’s how I think about it.

If you’re in the stage where you’re ready for feedback, there are options for you. If you’re a young writer (under 26), The Young Writers Initiative offers a free beta reading program where our dedicated volunteer beta readers will read and provide thorough and detailed feedback on your work. We also run a summer camp and workshop where you meet with other peers and develop a writing community that gives that real human feedback. We’re also gearing up to bring back our editing program by the beginning of 2026. There are so many opportunities for you to get your work read and out there.

Applications to be Beta Read are open for our next cycle! They close August 26th. If you’re interested in being a volunteer beta reader, stay on the lookout for applications dropping soon!

TYWI Beta Reading Service

What are your thoughts about using AI as an editor? Have you used it before? What do you think about it, and our points on its inherent flaws?

This post was written by Riya Cyriac, Executive Director of The Young Writers Initiative, writer, Human-AI Researcher, and product manager. Her undergraduate thesis examined how Generative AI designs are incompatible with the natural processes of writers, and how metrics are skewed towards AI adoption over writer augmentation. If you’re curious about her findings, you can read her thesis here.

By:

Riya