I’ve been using ChatGPT for over a year now, for all sorts of things.
I admit that I have used it to replace some of my own writing (I can hear many people tutting), mainly when I’ve had to write sales copy and podcast show notes.
I’ve also used it to help me come up with titles and domain names, write a section of the “bundle” I had to present as part of a law case (I’m not a lawyer!), write questions for course assessments, create adaptations of Shakespeare monologues etc, etc.
Working with this tool has rekindled my drive to write, and was a factor behind deciding to drop the client work with Virtual not Distant, to dedicate myself to writing and podcasting instead. (And before you ask, no, I don’t know how I’m going to make money out of that, but you know what, the world of work is fast evolving, and I’m nimble!)
After a few years of not releasing a non-fiction book, last year I published The Remote Worker’s Guide to Time Management, co-written with ChatGPT - against a lot of advice and commentary out there which advises against generating words with the tool instead of using it only as an aide.
(If you are not remotely interested in the subject (sorry, had to!), but are curious what the final content looks like, then message me and I’ll send you an e-copy.)
I am proud of the book. I had many reasons for co-writing it with GPT-4, and one of them was that I think it will be useful to others. But I’ve held back from promoting it too much, as this writing process still raises many negative emotions. I’ll let the dust settle while I get on with the next book.
I’ll write about that process another time, but for now, I want to share how I’ve used GPT-4 for feedback when writing my fiction.
Why I Co-Write Non-Fiction, but not Fiction with ChatGPT
A simple answer: the fiction it generates isn’t good enough.
While the non-fiction text it’s generated for me was more than acceptable, the prose I’ve seen from ChatGPT is terrible. It’s clichéd, and it reads like someone trying very hard to “write fiction well”.
Here’s an example of the bot re-writing one of my scenes to help me improve it. Here, the main character of my cozy mystery*, Mia Martinez, is talking to the police officer in charge of the investigation.
(*A cozy mystery is a crime novel where there is no blood, swearing or sex.)
“You’ve decided to delve into a murder investigation on your own?”
“Yes, I...” Mia hesitated, the weight of her decision suddenly pressing down on her. “After hearing whispers at a party post-murder, labeling me a suspect, I realized I needed to clear my name, or at least understand the narrative forming around me.”
DS Baker’s expression remained impassive, yet there was a hint of curiosity in her eyes.
“I can’t discuss specifics about suspects, Miss Martinez. You understand.”
When writing non-fiction, ChatGPT can write prose that is better than some self-help books out there. But not so much with dialogue and fiction. Or maybe I don’t know how to prompt it well.
BUT I have been using the tool for months now to help me improve my novel, and to help me find the motivation to tweak it, and tweak it, and tweak it until it sounds good enough to hire a proper editor to have a look.
It’s Not Human - and That’s the Point
I've wanted to write about this process for a while, but something I read today prompted me to get down to it.
I was reading a magazine about writing, featuring a letter to the editor explaining how someone in their writers’ group had showed them how they were obtaining feedback from ChatGPT. The rest of the group were outraged: why would you go to a bot for feedback, when you had a group of human writers to help you?
As someone who’s sought feedback from both, I am in a good position to comment on this.
Last summer, I took an online course, and while the content was average, the experience of having twelve other people to get feedback from, and to give feedback to, was invaluable.
It was a joy to read the work in progress of others, most of whom were more experienced writers than myself. In going through their work with a critical eye, I learned a lot about what was working in their writing and what I should do differently. When asking for feedback, I got many opinions about my writing - mainly about the content and characters.
I’d been working on the draft for the novel for about two years, making tweaks and corrections, but there’s nothing like knowing you’re going to have to put the piece in front of other writers to make it as good as you can, not just ok.
I spent hours working on scenes of about 800 words. Rewriting them, running them by Mollie, (as I called ChatGPT then) and when I thought I couldn’t improve them further, pasting them in the forum. For one of the scenes, I got a comment that said something like, “Well, if this is your first go at this, it’s very good.” (Before you think I was presenting myself as a better writer than I am, no writer would share their very first draft, we know that much, and having something ready to show doesn’t mean you’re anywhere near being done with it.)
To be honest, sometimes I felt like there was way too much encouragement and positive feedback coming my way, and not enough criticism. But the most valuable lessons I took from the course were:
I was developing a process for using ChatGPT to improve my writing
Working with ChatGPT was making me a better writer
In an ideal world, when we seek feedback from others, we’d get the feedback right when we needed it, from the right person and in a way that was useful to us.
With ChatGPT, I can get the feedback when I need it, it’s useful, and, even if it’s not always what I want, it doesn’t matter. I don’t need to thank the person for their time, or find a polite way of asking them to give me feedback again.
Something else I’ve noticed is that, when someone who rejects the use of ChatGPT for activities traditionally done by “humans”, they rarely mean “humans”: they mean “the best version of those humans”.
ChatGPT writes better than some writers and gives feedback better than most. It has its limitations - but so do humans. (I’ve seen something similar happening with the romanticising of the return to the office, as if before the pandemic, all office spaces were a joy to work in…)
The Feedback Process
Since I took the writing course, every morning I go back to my “first draft” to edit and rewrite one or two scenes, and I end the session with an interaction with ChatGPT. Through the session, I might interact with it if I’m stuck, but that’s material for another blog post.
I have one chat which I’ve been using since very early on, to ask for feedback. This means that ChatGPT kind-of knows what I’m talking about, recognises most of the characters in the story and sometimes even links the scenes to previous work. (OpenAI is working on giving ChatGPT a “memory” across your chats, but this hasn’t rolled out yet to common people like me.)
To wrap up this post, let me give you some examples of what these feedback interactions look like. My comments are in brackets and italics.
ChatGPT:
This is another compelling instalment! (Even if it might not be true, this is always nice to hear!) The description of Mia’s observations is interesting and her “murder books” add another layer to her character. Here are a few notes to consider:
Setting: This section gives us a nice look into Mia’s routine and thought processes. The café adds depth to the atmosphere, but the specifics about it seem slightly disconnected from the main action. While the café certainly sets a certain mood, perhaps you can make the details more relevant to Mia’s experience or thought process to make the setting feel more integrated.
“Murder Books” Context: The concept of the “murder books” is intriguing and makes Mia a more multi-faceted character. However, it comes quite suddenly. You might consider seeding the idea earlier in the novel so it doesn’t seem out of the blue. (Yeap, I do this a lot. I move on too quickly from one idea to another.)
Dialogue: Most of the dialogue is organic and adds to the characterization. The last part with Toby, however, feels slightly awkward, especially when he explains they’ve experimented with many blends. This seems like an odd thing to say to a customer, unless it’s a lead into further conversation. Perhaps it could feel more natural if Toby comments on her notebook or they have another topic to discuss.
Vocabulary: You may want to consider a more vivid verb than “met” in the sentence “Mia looked up and met the smile of a man...”. You could say something like “Mia looked up and was greeted by the smile of a man...” to add more active language. (When it offers things like this, they go straight in!)
Show, Don’t Tell: The line, “Pretending like it hadn’t happened, or ‘being professional’ about it was the best tactic,” tells the reader directly what Mia’s thinking. You might consider showing her thoughts instead, through her actions or a more indirect narration, for instance: “Mia steeled herself, pushing thoughts of the argument to the back of her mind. Professionalism was key now, a shield to the awkwardness.”
It always tells me to Show, Don’t Tell, but I’m ignoring it. I’ve read enough books to know that many successful writers, including many I like, let their characters be explicit. Anyway, I think the solution it suggests doesn’t solve the problem!
Here are some other bits of feedback, on other pieces. As you’ll see, you get all kinds of feedback, developmental, and line edits.
Character Descriptions: Your description of the young producer adds to the overall impression of the studio’s atmosphere and the types of people Mia interacts with. However, Mia’s musings on his age, appearance, and the way he wears his jeans might be seen as slightly judgmental, which could affect how the reader perceives her character. Unless Mia’s judgemental nature is an intentional part of her character, you might want to revise this part to ensure the tone aligns with the character you’re portraying.
Tone and Humor: The humor in this piece, particularly in Mia’s observations about the young producer, works well and helps to lighten the overall tone. This helps to create a contrast to the more serious and ominous elements of Mia’s story ideas.
Subtle Tension: The open fire exit door adds a subtle element of tension to the narrative. Consider highlighting this a bit more, as it’s a great tool for foreshadowing and could be developed into a larger plot point later in the story.
Every now and then, ChatGPT hits on something that’s been worrying me, so I follow it up. For example, regarding the last point on “subtle tension”:
Me:
Regarding point 6, you’re right, it’s foreshadowing a potential clue for later, that turns out to be a red herring. do you think that drawing more attention to it might be a bit too much for the reader, or does it need something more? (You’ll notice the typo, with a small “d” after a full stop. That’s another good thing with working with the bot rather than a human, it understands typos, misspellings and doesn’t bat a digital eyelid!)
ChatGPT:
Given that this clue is a red herring intended to mislead readers, you may not want to draw too much attention to it, or it might seem overly conspicuous. As it stands, the detail about the fire exit door being open is subtly presented, which seems appropriate.
However, to give this event a bit more significance (without making it too glaring), you could have Mia feel a brief moment of unease or confusion about the door being open. For instance, Mia could briefly ponder why the door might be open, given the studio’s usual attention to security, or she could feel a momentary chill as she notices the draft.
Remember that creating effective red herrings is all about balance. You want to give the clue enough weight so that it’s remembered, but not so much that it screams ‘look at me!’ to your readers. Therefore, a minor adjustment to enhance Mia’s reaction could be enough. (The bot does this often, trying to educate you as it goes along.)
Finally, consider calling back to this detail later when it becomes relevant again. The repetition will serve to remind readers of the event and enhance the sense of coherence in your story.
Final Words
I could go on and on about this, and indeed I will in consecutive posts.
(By the way, unsure of the use of “consecutive” here, I asked ChatGPT if the sentence was ok.
ChatGPT:
“Yes, your sentence is grammatically correct and clear in its intention. It effectively communicates that you have more to say on the subject and plan to do so in future posts. It’s a good way to engage your readers and keep them looking forward to more content. If you’re aiming for a slightly more formal or polished tone, you might consider:
“I have much more to share on this topic, and I will continue to do so in upcoming posts.”
But your original sentence has a conversational and inviting tone, which might be exactly what you’re aiming for depending on your audience and the context of your message.)
Final Words - Take 2
I could go on and on about this, and indeed I will in upcoming posts. But I hope this piece has shown you what people mean when they say that GenAI can “democratise” writing. For $20 a month, you can learn to write and learn about writing as you write. Of course, you need to have a sense of what good writing looks like in the first place, as the bot is not a writing expert (yet!).
So no, working with ChatGPT is not like working with a human. And that’s where most of its value lies.
For myself, using the tool will help me get my writing to the level where I can engage a human developmental editor, and make the most out of them. Although by the time I’m done with this draft, there might be an app for that!
By the way, this post was initially going to cover a range of ways in which I’m using ChatGPT for the cozy mystery, including developing plausible ways of committing a crime. But when I started writing about how to get feedback, I realised it needed a whole post to itself. If you want to know more about the process of writing fiction with ChatGPT (other GenAI tools available) click on the Subscribe button.
And let me know if you have any questions you’d like me to answer in future posts!