On Being Flagged
As 100% AI Generated
GPTZero says the essay I wrote about Lord of the Flies — about moral injury, about the boys who lost their scaffolding along with the adults, about the black-and-white film I watched alone as a boy because Mr. Rogers’ Neighborhood wasn’t real to me and Golding’s island was — GPTZero says it’s one hundred percent machine. Not mostly. Not “further review recommended.” One hundred percent.
Let me be precise about what stings here, because it’s specific, not generic.
It’s not vanity. I don’t need a certificate saying 100% human-made to know what I spent on that piece. I know where the idea came from — a boy in front of a television, working something out he didn’t have language for yet. I know how long it took to turn that memory into a coherent argument about attachment theory and scaffolding and Roger’s stones deliberately missing their mark. That labor doesn’t need a machine’s approval to have happened. But I still feel the sting, because the accusation isn’t neutral — it’s a claim about where the essay came from, and it’s a wrong one, made with total confidence, about the one piece of writing I can trace back to an actual afternoon of my own childhood.
I used Claude to clarify sentences. To catch a dangling modifier, to ask whether a paragraph is doing what I think it’s doing. That’s the whole of it. The thinking — the memory of that television, the years of clinical practice that let me see moral injury in Roger and Jack and Ralph where other readings only saw savagery — that’s mine, arrived at slowly, the way anything true arrives. A detector cannot see any of that. It isn’t built to. It was built to find rhythm that resembles other rhythm, and mine, apparently, resembles something it’s decided to call synthetic.
I don’t have a neat place to put this feeling. I could say the tool is wrong and move on, but that undersells it — being told your voice reads as no one’s voice in particular is its own small grief, adjacent to but not the same as being told you’re lying or you’re a fake. I know the difference between being disbelieved and being erased, and this is closer to the second. Not “we don’t trust your account.” More like: there was no one home to have an account.
There’s an irony in this essay specifically. The piece argues that the boys on that island didn’t turn savage because savagery is what’s underneath — they turned savage because the scaffolding that should have shaped their moral sense was ripped away too soon, and nothing was there to catch them. A detector that can’t tell a lived memory from a generated one is its own small failure of scaffolding: a system built to hold a line between human and machine, deployed with total confidence, that collapses the moment it meets something it wasn’t built to do. It doesn’t know the difference. It just says one hundred percent, and moves on. Fuck that.
I don’t have an ending for this one either. I’m always suspicious of the ones that resolve. But I notice what’s happening: the tool built to catch the absence of a mind is, in practice, catching the presence of one — mistaking a lifetime of noticing patterns like this one for evidence that no one did the noticing at all.
I ran this essay through GPTZero before publishing it. It came back 86%. I’m including that number here on purpose, and I’m not explaining it.
Letters from a Psychotherapist is a reader-supported, ad-free publication. When you subscribe, you’re directly supporting this work and helping keep it independent and accessible.
I’m grateful you’re here.


Bottom line: AI could not have written that piece without your substantial and meaningful assistance. Keep doing you, using any tools you like!
Do I understand you posted original article to substack first. Later you ran it through the AI whatever?
Does AI assume that if you get 100% that means no way a mere human could come so close to perfect???
Or does it mean AI did a massive clean up job and you're a complete dolt??
Another useless statistic!!