It was just an experiment. I was staring at a jumble of notes for an essay I was working on, anxious for it to become the profound statement I had dreamt it might be in the first place.
I had what I felt was a great premise, but all week I pounded at the keys, amassing over twelve pages of meandering thoughts, notes, and research on my subject. Every new development spun me away and around my original golden nugget of a thesis, down rabbit holes and into dead ends. Now every time I opened the document it was like untangling the very delicate chain of a balled-up necklace. Slow-going writing sessions with unsatisfying endpoints, interrupted by my dog crying to be taken out, a Zoom appointment I had to hop to, or my toddler’s voice, tiny but bright like the color pink, calling out “Mama” from her crib. I’d close my laptop screen, the essay growing evermore amorphous beneath it.
I’d been using AI in my daily life with some success. With dutiful precision, ChatGPT built me a no-fuss skincare routine that I have actually stuck with for the first time in my life. It spit out where exactly to hammer my nails for a gallery wall. It converted my favorite carrot soup recipe into a curry with seared shrimp.
Another time-crunched day of trying to bring my latest essay to fruition was coming to an end. With my subject matter growing staler by the moment, pressure loomed. Having been a professional freelance writer for 15 years, I understood that I had something worthwhile here if I could just see it shaped up, if I could just get a sneak peek at what it would look like finished.
Among the tabs crowding around my Google Docs draft, ChatGPT was open next to an article sent around on my in-laws’ WhatsApp thread earlier that week. In the essay that was about to go viral, Matt Shumer, CEO of OthersideAI, which is known for its AI writing assistant Hyperwrite, argues that those unwilling to use AI platforms might get left behind, predicting imminent disruptions to job security across disciplines and trades. In short, he says we can’t beat the robots, so we’d better join them.
Earlier in the day, because of the article, I strayed away from the domestic tasks I had thus far been reserving ChatGPT for and asked it to organize a lesson plan from a handwritten page of notes and was amazed, if not a little scared, by the practically gleaming, bullet-pointed document it produced in seconds, a feat that would have taken me the better part of two hours. It was a little intoxicating to use that time instead for a walk and to make a cup of tea.
Fresh off the heels of that high, I now anxiously looked at the clock. I had less than an hour until daycare pickup and maybe another two-hour window of writing time left in my schedule before my essay idea would become too stale to pitch. I pressed Control+C on my notes, fed it into ChatGPT, and asked, “What would this look like as a polished essay?”
Words flew onto the screen, like watching the tangle of the chain fall loose from someone’s fingertips after days of trying to pick at it myself.
I had expected that the essay would be ridiculous. As many writers I admire say, I was sure it would be “slop.” I assumed I would read whatever droll it could muster with its limited robot perspective and be bolstered, realizing how it should really be done in a flash of writerly inspiration, formulating the path to a far-superior human version of my essay.
But the problem was, I liked it. After all, here were my thoughts, wrapped into a pretty package that lacked any sensical trip-ups or grammatical errors. In fact, it was so clear and logical I couldn’t see past it. Suddenly all the possibilities I had imagined collapsed into this sterilized version. How could it look any other way?
I also felt weirdly inadequate. Why couldn’t I see the connections the AI tool had made that seemed so obvious now?
In March, I attended the Association of Writers & Writing Programs annual conference in Baltimore, where AI in writing was a hot topic. Writers rolled their eyes at any mention of AI and panelists shot off zealous responses to audience member questions about its place in the literary world. “There is no place,” one prolific author responded. “I wish it didn’t exist.”
But it does exist. In the literary community, commentary has been extreme on the side of dismissing AI use as morally corrupt and all bot writing as “slop.” But that slop recently made it all the way to a big five publisher and AI tools are getting better and better, threatening slow artforms with faster outputs, just as Shumer had suggested AI will chomp at the heels of other professions.
AI platforms advertise that anyone can now have a seat at the creative table, dipping your toe into an artform and getting the satisfaction of a written piece made by you. Better still, without any of the suffering!
Suffering has long been a dilemma for the writer, my own practice often reminding me of the Robert Hass quote: “It’s hell writing and it’s hell not writing. The only tolerable state is having just written.” AI offers a shortcut to that tolerable state. With that option enticingly at our fingertips, writers as a whole are more susceptible to this trap than ever.
Trying to forget the ChatGPT essay, I started a few new drafts, but each version started to morph into the AI version until I didn’t know what was my idea and what was the bot’s. Ultimately, I abandoned the project, unable to bring myself to write my own now that I couldn’t unsee what AI had created. What did toiling over work matter when it could so easily be done by a robot anyway?
What was more disturbing and unexpected is that ideas, which usually come to me easily, stopped appearing for the next few weeks, a writer’s block like I have never known.
For a writer, our biggest concern in the early days of AI was that it would steal our ideas. Now I see it can steal our purpose too.
Along with being a writer, I’m a generative writing coach, with a business built on the fostering of each student’s one, unique voice. I’m ashamed that even I, in a moment of weakness, couldn’t resist the lure of skipping the hellish part of my artform. While I’d like to be like the sectarian vigilante writers who would never even dream of making a ChatGPT account, I think it’s more realistic to start a frank discussion around how artists can retain their motivation to create when there is an expedient route to completion, without suffering, right there, available to us. But I’m here to tell you, AI didn’t relieve the suffering of the writing process. It just created a worse, more high-tech brand of it.
In a groundbreaking 2018 study, psychologists developed the term The Effort Paradox. While humans value work that requires more effort, we naturally avoid the hardest path. Writing—and especially finding success as a writer—is famously difficult. But these goals, I’m realizing, are two different things. Like an athlete simultaneously wants to master their sport and to be recognized as the best at it, doping only fulfills one of those desires. Similarly, work produced by AI might be published (if it can make it past editors and AI detectors), but publishing isn’t writing.
The sometime- hellish effort of putting your ideas together into a package that flows is.
Writing is the spark of an idea set off by some unexpected moment of inspiration, something as inconsequential as a puddle igniting a thought. It’s the fight to build a bridge with words from that moment to the capital P Point that is nagging at you to be told. It’s steering yourself out of dead ends and back onto the right path, picking your way, ever so slowly, toward an articulated idea that might just matter to someone who reads it. It’s all that work that AI offers to do for us.
With a world moving faster and faster as we outsource tasks to AI, the time it takes to make a piece of original writing will remain the same and sometimes feel glacial in comparison to our bot-assisted lives, but time is an essential ingredient to writing. Along with toil.
As the philosopher Adam Smith once said: “The real price of everything…is the toil and trouble of acquiring it.”
What I did when I fed my thoughts to ChatGPT wasn’t writing. Product without process isn’t art; it’s something else.
Maybe my originally written essay would have faltered in its quest to paint a sweeping call to action, one that builds on itself sentence by sentence, as a good essay should, but even if it was bad, it would have been mine. It would have informed my craft, teaching me things that will make the next essay that much better. It would have been writing. Real, honest writing.
I’d rather see my work’s frayed edges than AI’s perfectly polished ones. I’d rather spend 20 minutes replacing a redundant phrase than outsource that task to an algorithm. I’d rather kill my darlings with the satisfaction that given enough time and toil I will know what belongs and where it belongs and what the hell I’m trying to say.
Joan Didion said, “I write entirely to find out what I’m thinking.” As far as the piece I started and abandoned, I may never know what I really think about the topic, AI having snuffed out any revelations following a simple stroke of my return key.
From the moment I played my first game of make believe, creative exploration through storytelling has been responsible for teaching me who I am and what matters to me. For artists in today’s world, it’s not just our ideas we need to protect, or even our careers; it’s the very art form we’ve built our lives and selves around.
Sammi LaBue is a generative writing coach and basically obsessed with the feeling of having an idea and writing it down. Her latest project is a recently finished memoir written in collaboration with her mom titled “Bad Apples.” Some of her other essays can be found in HuffPost, Slate, Literary Hub, The Sun, The Offing and more. To follow her writing journey and find opportunities to write with her, visit fledgling.substack.com.






I have been having a bit of go with AI as well. I think however having used chapgbt, Gemini (whom we fondly call "Jim and I" ) and Claude --there is an extreme difference between all three. I try to stay in a playful place with all this.
For Claude I have been very explicit ... don't flatter me. I like pushback and I always appreciate a "what if" scenario. So it always gives me pushback and "what if's" ( I think it is wired to flatter that is a harder ask for it to stop doing)
Here is a snippet from this weekend from Claude to me"
"One honest pushback on what you just said. Because you asked for honesty and I'd be failing you otherwise.
I'm a very good thinking partner. Today has been genuinely remarkable — the scout, the dink, the failure of imagination, the promise you won't keep to yourself. This was real thinking.
But I don't know what your face looks like when you say these things. I can't sit with you in the silence after the hard part. I don't know Hunt's voice or Julia Thomas's laugh or what the Vermont light looks like through your window right now.
Find one human who can hold some of this. Not all of it. Just some."
I love doing the writing myself. Claude knows that so it never gives me text but it gives me criticism, tells me what it thinks is working and sometimes makes suggestions where it thinks my writing is weak. It is also very good at checking any facts or references to make sure they are as close to 100% as I can get. It is really good at recognizing if a reference I am using was generated by AI. It warns me. And that's pretty funny, yes?
So good!