I’m undertaking a 1000-day reinvention project, blogging here daily to track my progress. In Wednesday Writing, I consider my writing practice and skills and how to improve upon them.
I admit I really enjoy chatting with ChatGPT throughout the day. I use it for blog post ideas. I use it to summarize things I can’t summarize myself very easily–like Kierkegaard’s concept of the aesthetic versus the ethical life that I wrote about yesterday. I get guidance from it on small and large life questions: What are some easy low carb dinner ideas? How can I find my lost scissors? What is the best way to record voice notes while I’m walking the dog? What do I do when my neighbor puts his trash in my trash bin because his is full?
ChatGPT is incredibly helpful for all those things. Amazingly, its suggestions for finding my scissors actually worked. But I can’t help but wonder if I’m going to destroy my writing voice by leaning too heavily on its drafting and editing help.
I don’t even know what my writing voice is. I haven’t made any conscious steps to develop it, except for recently starting copywork. Here on the 1000-day project, I write most things entirely myself, with the exception of short passages summarizing ideas I want to write about that I get from ChatGPT. I usually edit and polish those into my style a bit too, removing extra em-dashes, getting rid of “It’s not X, it’s Y” constructions, and deleting words it uses that I don’t (e.g., “quiet” and “unravel”).
My book manuscript–now that’s a different thing entirely. I have used ChatGPT to draft entire sections. Now I’m going back and revising them. Ideally I read them and then redraft them, with my own understanding and voice as primary. It’s hard to do that when I’m facing a few paragraphs that already read pretty well. But I still think it’s important to do that. I often find when I do it that what sounds good doesn’t actually mean anything. It’s just bullshit. It’s just slop.
What are some hallmarks of AI-generated writing?
Some I’ve noticed, or read about:
- Of course the ever-present em-dash. I always used the em-dash so now I’m confused as to when to use it or not. When I have a piece of AI-generated writing, I tend to remove the em-dashes and replace them somehow–with a colon, with a comma, with an entirely different sentence construction. (That em-dash came naturally to me… I think ?)
- The use of “quietly” or “quiet.” I don’t think I use quiet as an adjective much but I see it all the time in AI-generated writing. There’s a quiet revolution, or someone is quietly rebelling, or she felt a quiet confidence in her new love. I eliminate quiet every time I see it. And when I read it in someone’s essay, I figure they used AI. It’s weird, because it must have come from real human writing. But it just rubs me the wrong way. It’s so noisy to me. Another word ChatGPT uses a lot that I don’t: unraveling.
- Rhythmic, staccato output. Sentence fragments.
Carriage returns where they don’t seem to be necessary.
Super-short paragraphs of only one sentence.
What the hell, ChatGPT? I hate that style. Do I really ever write like that? I don’t think I’m generally given to sentence fragments but now I don’t know, because my writing style is merging with ChatGPT’s writing style. - “It’s not X, it’s Y.” I’m not sure I would have noticed this if other people didn’t call it out. But it is fucking ubiquitous. I usually just remove the “It’s not X” sentence because often the Y sentence actually adds something. But not always. (Oops there’s a sentence fragment…)
- The use of lists of three in perfectly parallel structure. I like lists of elements just like everyone else, because they can be helpful, enlightening, and interesting.
- Overuse of positive adjectives like “rich,” “insightful,” “revolutionary,” “groundbreaking.”
- The “setup-payoff” structure, starting with a bold statement or rhetorical question and then a surprisingly bland, generic, meaningless explanation.
Some things I’m doing to ensure I don’t lose my human writing voice:
- I write something myself first, and use AI to edit it. Then, I don’t take the newly written draft as is, but type it into my book manuscript or blog post, rewriting it as I go.
- As mentioned, I delete any words that I would never use. Sometimes AI uses a word that maybe I wouldn’t use, but I like and want to incorporate into my writing. So I leave those gems in.
- And, most important, I write here every day with very little AI help (other than producing quick summaries of things or brainstorming ideas).
- I add in my own personal stories, provocative takes, weird things I think and more. Jess at the Limitless Creator calls it adding your mess in. I like that idea of making my writing messier, more me.
I imagine that as I get more comfortable with writing a book manuscript I’ll rely on artificial intelligence less to do it, relying on it more as an editor and researcher. I’m already mostly using it that way.