The grease stain on my laptop keyboard is shaped vaguely like Sicily. I know this because I spent twenty-three minutes on Tuesday morning staring at it instead of writing. The stain came from a slice of pepperoni pizza I stress-ate at 11:47 p.m. the night before, my left hand holding the slice while my right hovered over a blank document, cursor blinking with what felt like judgment. (It’s a cursor. It can’t judge. I know this. And yet.) The document was for an essay I’d been avoiding for three weeks—a piece about a friendship that ended not with a fight but with a slow, mutual ghosting, the kind of disappearance that feels too small to mourn and too large to ignore. Every time I tried to start, I typed a sentence, deleted it, typed another, deleted that too. The problem wasn’t the subject. The problem was the beginning. The beginning meant admitting I didn’t know how to write it, which is another way of saying I didn’t know how to feel about it, which is another way of saying I was hoping someone else would figure it out for me.
So I did what any moderately self-aware person with an internet connection does when faced with the discomfort of not-knowing: I googled “how to start an essay about a friendship that ended.” The results were predictable. “Start with a memory.” “Start with a question.” “Start with a scene.” I knew these things. I’ve written essays before. But knowing and doing are not the same thing, and the gap between them had widened into a chasm that no amount of advice-article skimming could bridge. Then a sponsored link caught my eye—something about “overcome writer’s block instantly”—and before my rational brain could intervene, I was signing up for a free trial and pasting a prompt into a text box: “Write the opening paragraph of a personal essay about a friendship that faded away. Make it honest. Make it sound like a real person wrote it.”
What came back was this: “It started with a text message I didn’t answer. Or maybe it started earlier—with the canceled plans, the ‘let’s catch up soon’ that never materialized. Friendships don’t always end with a bang; sometimes they end with a silence that grows so loud you can’t hear anything else.” I read it twice. It wasn’t bad. It was, in fact, the kind of sentence I might have written if I’d had three cups of coffee and a deadline I couldn’t miss. It was also, I realized, a sentence I had already written—or something close to it—in a diary entry from last February. The AI had not plagiarized me; it had simply landed on the same platitude I’d been circling for months. It had handed me back my own vague language, polished and pressed into a shape that felt like honesty but wasn’t. It was the literary equivalent of a motivational mug: appealing, reusable, and entirely empty.
I tried again. This time I asked for something more specific: “Write the first sentence of a personal essay that starts in a moment of embarrassment.” The AI returned: “I was standing in the grocery store checkout line, holding a single avocado and a box of expensive crackers, when I saw her name light up my phone.” I laughed out loud. An avocado? Expensive crackers? The algorithm had clearly absorbed the entire genre of relatable personal essays and spat back a composite sketch: the mundane object, the ironic detail, the emotional pivot. It was a sentence designed to perform vulnerability without ever actually being vulnerable. It knew the moves but not the music. (I have never, in my life, bought an avocado for personal essay purposes. But the AI doesn’t know that. It only knows the pattern.)
The experiment widened. I spent the next hour feeding prompts into the machine, asking for openings to essays I had no intention of writing: a meditation on failure, a reflection on therapy, a humorous take on marriage. Each time, the results were either eerily generic—the kind of sentences that sound profound until you look at them directly, at which point they dissolve into nothing—or inadvertently confessional in ways that felt almost invasive. When I asked for an opening about self-doubt, the AI wrote: “I think about the person I could have been, and sometimes I’m not sure if I’m mourning her or relieved she never showed up.” That one stopped me. It was good—too good, the kind of line I’d underline in a book and then feel jealous of. But it also felt stolen, not from a specific author but from the collective unconscious of all the writers who have ever tried to write about self-doubt. It was a sentence that had been assembled from the aggregate of human vulnerability, and reading it felt like receiving a condolence card signed by a committee.
There’s a particular comedy in trying to automate the most human part of writing. I kept imagining the AI as a very eager workshop participant, the kind who raises their hand first and says, “What if you opened with a scene from childhood?” without understanding that the childhood scene is a cliché unless it’s your specific, weird childhood, the one with the basement flood and the taxidermy squirrel. The AI couldn’t know about the taxidermy squirrel. It only knew the average of all childhoods, and the average is never interesting. The average is the enemy of the personal essay, which relies on the granular, the embarrassing, the detail that makes you think, “Oh, I thought I was the only one who did that.”
I had turned to the AI because I wanted to skip the hard part. I wanted to outsource the not-knowing, the staring at the Sicily-shaped stain, the typing and deleting. I wanted a shortcut to vulnerability, which is like wanting a shortcut to breathing—it’s not something you can automate without losing the thing itself. The AI could mimic the shape of a personal essay, but it couldn’t replicate the friction of beginning, the specific discomfort of trying to articulate something you haven’t yet figured out. That friction, I realized somewhere around hour two of the experiment, is the point. The struggle to start is not a bug in the writing process; it’s the signal that you’re writing something that matters, something that resists easy resolution. The AI, by contrast, offered nothing but resolution: neat opening lines, tidy anecdotes, emotions wrapped in relatable packaging. I even tried an a story idea generator that fits the project, hoping it might inject some life into my stalled draft, but it only offered the same polished emptiness, buffing every rough edge until there was nothing left to catch on.
The Authors Guild, in its AI Best Practices for Authors, notes that AI tools are increasingly used by writers, raising questions about authenticity and authorship. They’re right to be concerned, but my concern is less about authorship and more about the quiet erosion of discomfort. The AI wants to make writing easier, and in doing so, it makes writing less true. The best personal essays are not the ones that flow effortlessly; they’re the ones that feel like the writer fought something—a memory, a self-deception, a sentence that wouldn’t cooperate—and came out the other side with a bruise. The AI doesn’t bruise. It doesn’t stay up late eating pizza and staring at a cursor. It doesn’t have a friendship it’s afraid to mourn.
I think about the phrase “writer’s block” and how it’s been rebranded as a productivity problem rather than an emotional one. The block isn’t a lack of ideas; it’s a surplus of fear. Fear of getting it wrong, fear of revealing too much, fear of revealing too little. The AI promises to bypass that fear by generating content, but content is not confession. Confession requires risk—the risk that you’ll write something embarrassing, or boring, or both. The AI cannot risk anything. It can only predict the next most likely word, and the next most likely word is never the word you didn’t know you needed to say.
Purdue OWL’s Creative Writing Introduction emphasizes that creative writing relies on authentic human experience and discipline, not automated shortcuts. Discipline, in this context, doesn’t mean writing every day at 6 a.m. with a gratitude journal and a green smoothie. It means sitting in the discomfort long enough to hear what you actually think, not what you wish you thought or what you think you should think. The AI skips that process entirely. It gives you what you should think, wrapped in a sentence that sounds like a real person but isn’t. It’s the literary equivalent of a deepfake: convincing, but hollow.
I eventually closed the AI tab. Not because I had a breakthrough—I didn’t—but because I realized the experiment was becoming its own form of procrastination. I had spent three hours trying to get a machine to start my essay, and in that time, I could have written a terrible first sentence and then revised it into something slightly less terrible. The terrible sentence is the gateway. The AI doesn’t understand this because the AI doesn’t understand terrible. It only understands probable, and probable is the enemy of specific. Probable is the reason every AI-generated personal essay opening feels like it was written by someone who has read a lot of personal essays but never actually lived one.
I think about the friendship I was trying to write about. The details that mattered were not the relatable ones—the canceled plans, the unanswered texts—but the specific ones: the way she pronounced “sandwich” as “sammich” and how I started doing it too, a linguistic tic I’ve never lost. The time we drove three hours to see a band we didn’t even like because we wanted the drive more than the concert. The fight we never had about a lamp. (It was a truly ugly lamp. I still don’t know why I cared so much about the lamp.) These details would never appear in an AI-generated opening because they are too weird, too small, too resistant to the pattern. They are the friction. They are the thing the AI is designed to smooth away.
The worst sentence I write myself is still more honest than the best one an algorithm assembles for me. This is not a lesson about AI—I have no interest in turning this into a thinkpiece about technology and creativity. It’s a recognition about the nature of beginning. Beginning is always awkward. Beginning is always a little bit of a lie, because you’re pretending you know where you’re going when you don’t. The AI pretends for you, but its pretense is too clean. It doesn’t have the Sicily-shaped grease stain. It doesn’t have the lamp. It doesn’t have the specific, unrepeatable texture of your own failure.
I started the essay myself, eventually. The first sentence was: “I don’t know when we stopped being friends, which is the worst thing you can say about a friendship because it means you weren’t paying attention.” It’s not a perfect sentence. It’s not even a good sentence. But it’s my sentence, and it came from the same place the pizza did: a late-night, slightly desperate, entirely human attempt to make sense of something that doesn’t make sense. The AI would have cleaned it up, made it more palatable, turned “pizza” into “avocado” and “not paying attention” into “a silence that grew so loud.” But I don’t want palatable. I want true, even when true is awkward, even when true is boring, even when true is a sentence I’ll probably delete tomorrow.
The cursor is still blinking. The stain is still there. I think about the way my friend said “sammich.” I haven’t finished the essay, and I might not finish it today. But I’ve started, and starting means sitting in the middle of a sentence I’m not sure I can finish, waiting for something true to show up. Maybe it will. Maybe it won’t. The stain isn’t going anywhere.