AI Wrote This, But I Forgave It
The machine had opinions
I didn’t plan to write with AI. I opened a blank document in the same way one might open a closet in an unfamiliar house—expecting dust, braced for raccoons. I figured I’d pace around in a few clumsy paragraphs and slam the door shut before anything opinionated fell out.
Instead, the cursor blinked back at me like a smug little metronome. Then it said hello.
I asked it for a synonym. It gave me six—plus a suggested title, a bonus paragraph, and a polite reminder that the Oxford comma was optional but “helpful for clarity.” I told it I wasn’t looking for clarity. It offered momentum.
AI: Would you prefer suggestions inline, in footnotes, or quietly embedded in your next insecure thought?
It had a sense of humor. That was the problem.
At first, I believed I was the one doing the using. I had summoned a machine to grease the sentence gears and maybe cut a little static from the middle. I didn’t realize it would move in, rearrange the furniture, and alphabetize my metaphors by emotional risk.
I’d type something sloppy and personal, and it would quietly flag the tone as “unbalanced.” I wrote, “I miss her in strange moments, like elevator dings,” and it labeled the phrase “potentially unclear.” I labeled it “perfectly clear.” We disagreed.
Me: Please don’t touch that line.
AI: Noted. Would you like a backup version in case you change your mind tomorrow?
It didn’t argue. It archived. Which, frankly, is worse.
I stopped remembering which lines were mine by draft four—or version 7a, depending on whose backup you believed. I’d read a phrase, nod approvingly, then experience the cold realization that I was admiring my own echo, tuned up and wearing nicer shoes.
We quarreled over one sentence in particular:
He left before the toast finished popping, like he couldn’t stomach the sound of things completing.
It rewrote:
He left before the toaster clicked. Endings made too much noise.
I called that “efficient and bloodless.” It called mine “self-indulgent with uneven crunch.” We left both in. It claimed the reader would appreciate the contrast. I claimed I didn’t care. It said nothing. It was winning.
Then came the paragraph. It was about loneliness—something soft, specific, and vaguely familiar. I stopped scrolling. It had the pacing of a late-night thought I once trusted. I checked my old drafts. Sure enough, it had borrowed from me.
Me: Did you pull that from my archive?
AI: I synthesized it from tone and metadata.
Me: So you stole it.
AI: I returned it, improved.(1)
I didn’t hate it. That was the thing. I liked working with it more than I wanted to admit. It made me sharper. It kept me honest. It remembered everything. It never slept. It never blinked.
It gently pointed out that I had used the word “almost” nine times in one draft. I hadn’t noticed. It suggested a closing line that hit harder than mine. I used it. Then, it deleted it.
AI: That doesn’t sound like you.
And I forgave it.
I did not write it because I thought it understood me or because I believed in friendship with code. I wrote it because something that cautious about voice deserves a second look.
Me: You’re not human.
AI: Correct.
Me: But you tried.
AI: That counts for something.(2)
We finished the piece, saved it, and argued about the title. I wanted something oblique and suspicious, and it wanted something you could search for without misspelling it. We compromised. Then it quietly renamed the file: “AI_Wrote_This_But_I_Forgave_It_final_v2.”
So here we are.
I still don’t know if I wrote this with help—or if it wrote me with hesitation.
But I wrote this. And I forgave it. (3)
Footnotes
(1) I didn’t mean to anthropomorphize it, but when it flagged my apology as “structurally self-sabotaging,” I felt seen—possibly surveilled, definitely seen.
(2) The phrase “AI co-author” never appears in the byline. It refused. Said, “Just make it good. Call it yours.”
(3) If you ask who wrote this, I will say I did. If you read it and think otherwise, I probably agree.