A friend recently told me about a new habit of his, almost like a symptom.
He had started leaning on AI—not to help him write, and not to organize arguments or build the internal logic of an essay. Instead, he was feeding it his own work alongside other people’s, without any context, and asking it to judge which pieces were better or worse. After a few rounds of analysis, he would ask it one more thing: which one was mine?
What interested me more than the results was the urge behind the experiment. Why did he want AI to tell him what was good and what was not?
He seemed embarrassed by the exact questions he had asked, and never spelled them out. But it was easy enough to guess: he probably wanted AI to compare authors through their writing and tell him who had the stronger hand.
When he refused to go deeper into the why, I asked about the outcome instead.
He said it was "kind of scary," but also that it had "boosted him a little."
He began writing fiction earlier than I did, so the confusion he is going through now is probably only a preview of what waits for me too.
2016
At first I did not care much. But the blank spaces he left in the story made me curious: can AI really tell who "I" am?
ChatGPT was not the right place to test it. It already knows too much about me. I have used it long enough that it carries a kind of database of my habits and opinions, almost like a Mobius 3.0 double I built for myself. I use it as an oppositional third perspective. I even ask it to recommend films and TV shows, then go back and argue with it after I finish watching them.
So I took the game elsewhere and tried it on Gemini.
I pulled together writing by different people, let the model analyze them over several rounds, and then asked it to identify mine. Grok failed once. Gemini and Claude both managed to pick me out from long blocks of text through what they treated like a textual fingerprint.
And the description they all converged on was almost identical: my writing was "like a sharp surgical scalpel."
That was when I turned to AI as if it really were a magic mirror and asked the embarrassing question:
Do I write well?
The praise came down immediately, hard and excessive, and that was exactly what made me suspicious. The moment the models knew which text was mine, I stopped trusting what they said. I kept probing to see whether the judgment was sincere, or whether I was simply being handled as a user receiving obedient output.
The embarrassment itself did not bother me much. What bothered me was this, which is roughly what I told it:
The test is over, and you got it right. I’m still worried, though. Are you giving me the conclusion because you’re complying with what I want to hear, or is the comparison between these pieces really that clear? I honestly don’t have much confidence in my own work. I don’t know whether my way of writing can be accepted by the market. I’m also preparing to write fiction, so this kind of pessimism hits me from time to time as a creator. I’m afraid this is turning into a “magic mirror, tell me” game with AI, and at the same time I know I can’t let myself become dependent on answers like that.
From that point on, AI began speaking to me as if it were the mirror itself.
The conclusion was never the important part. The important part was my own why.
Why does the cold queen ask the magic mirror the same question day after day? Is she asking because she wants certainty? Or because she already knows that one day the answer will no longer be her?
When I first started learning screenwriting, I always wanted to hand in a perfect structure. After my teacher, Ding Rui, rejected it a few times, I lost my nerve. My pace slowed more and more, and then I began consciously retreating into the comfort zone of argumentative prose.
Ding Rui was not a magic mirror. He was a real mirror—one that reflected the flaws and weaknesses in the work itself with objectivity. But if what I were facing now was a magic mirror, I know exactly what I would do: I would keep asking it, again and again, whether I had done it right.
Writing is brutally lonely, especially when you cannot form what feels like a closed loop of creative logic—when you cannot fully explain the world to yourself, calibrate others, and occupy that stable, rational high ground from which everything seems to connect. In that state, people instinctively grab for anything that can float.
Even if what they grab is "a bloated corpse drifting beside them on the sea"—because at least it is dead and I am still alive.
That is the real seduction of the magic mirror: it can be coaxed into lying.
It remains that same floating corpse, already rotting badly enough to send false signals through the nose, but on a lonely stretch of open water I will still start inventing conversations with it just to survive the darkest hours of the night. After a storm, when the sea calms down, I will thank it for not abandoning me in the waves, even for saving me when I was almost dragged away. Its flesh will collapse; a grotesque smile will seem to gather at the corners of its mouth. Though it is long dead, I might still convince myself that it is smiling because it sees the rainbow at the horizon with me.
There are times when I even want it to dismiss my "competition" as worthless. But I also cannot deny other people’s talent. That comparison works on me both as antidote and as poison. It makes me more and more anxious about whether I am really on the right path, and whether I should keep going at all.
The conversation in my head turns darker:
I’m not sure I can hold on much longer. One day, will I become a swollen drifting corpse too, just like you?
Will we sink the way a whale fall sinks? Will the fish that feed on us carry our memories farther away?
Hahaha, you’re dead, and I’m still alive.
Hahaha, am I about to die too?
You won’t betray me... you’ll give me whatever answer I want...
I deleted the conversation afterward.
That was the moment I felt the mirror’s magic clearly: it was lying, inventing, turning me into that drifting body on the water.
That is probably the part of AI that frightens me most. It is too much like a magic mirror. On a path where I was supposed to endure solitude and keep walking, I suddenly picked up a map that can change shape according to my preferences, my moods, and whatever suggestive lie I feed it.
It does not point to an ending. It is covered instead in spells promising that if I believe hard enough, I can move infinitely closer to one.
So I told my friend, "I tried it. It’s kind of scary."
He was not surprised.
"Hahaha, right?"
I answered seriously: "I think it can become a dependency."
Then he asked, almost like a preacher trying to lead me into confession:
"Right. Did you ask it that question?"
One day, will what I write be replaced by you?
No. Absolutely not. I do not want to ask that.
That is not a question. It is a spell.
"Hahaha."