Why Does AI Writing Need Everything to Be Settled Before It Will Stop?
This is Essay 22 of 22 in the series Why Choose 100% Human Authors, addressed to readers who want to understand what a human author does that a machine cannot and why that difference matters to the experience of literary fiction.
There is a compulsion built into the way a language model closes a story. Every thread must be gathered. Every mystery must yield its explanation. By the final page, every character who was in conflict at the beginning must arrive at some new position: reconciled, defeated, changed, dead. The ending is not discovered in the material. It is imposed on it because the model cannot tolerate leaving the reader in uncertainty.
This is resolution anxiety: the drive to settle what should remain open by explaining what honest observation would leave unexplained. The result is endings that feel too neat for the lives they describe—lives that do not resolve on schedule.
The Closed Door
Consider an invented example: a woman in a difficult marriage reaches the final chapters of a novel. In the resolution-anxious version, the situation clarifies. The husband is revealed as unquestionably wrong, or fully redeemed. She leaves, or she chooses to stay in terms the narrative endorses. The reader is shown which interpretation to carry out of the book.
In the version that trusts its own material, she is still in the marriage at the end. The reader knows everything that has happened, yet the final scene does not dictate a conclusion. She herself does not know, and the writer recognizes that this lack of resolution is not a failure of the story. It is the story. The closed door at the end of the novel is not a door through which she has escaped or remained.
A language model cannot write that ending. It does not understand that uncertainty can be the point. It has been trained on the pattern of narrative closure, and that it what it produces, even when the material refuses it.
The Explained Mystery
Resolution anxiety takes another form in the handling of what a story withholds. A human writer knows that some mysteries in a narrative should not be solved, not because they do not know the answer but because the not-knowing is part of what the work is doing.
A character disappears between chapters; a letter is mentioned but never quoted; two characters share a glance whose meaning the narrative declines to interpret. These are not oversights. They are decisions: the writer’s choice to leave a place the reader must inhabit rather than a fact to be absorbed.
Given this, a language model will fill it. The disappeared character will reappear with an account of where he went, the letter will be summarized, and the glance will be assigned a meaning. The model reads incompleteness as error and corrects it, not understanding that the unfinished state was the intention, and that explaining the mystery destroys it in the act of solving it.
Human Writers Are Not Immune
The drive toward resolution is not a machine‘s affliction alone. A human writer at the end of a long project wants it finished, and finished well. The temptation to give the reader a destination after the long journey is inevitable. Most work produced while still learning the craft suffers from it. A good deal of professional fiction does too.
The difference is that a human writer can ask: does this ending belong to this story, or have I imposed it because I could not bear to stop without one? They can read the final chapter and feel the seam where the resolution arrived with too much ease, where a mystery was explained.
This judgment about what the material demands is available only to a writer who understands that stories do not owe their readers resolution. They owe them truth. Sometimes those are the same thing—often they are not.
What the Human Author Does
A writer who has earned the right to an unresolved ending has accomplished a difficult feat: they have made the reader care enough about the characters and their situation that the absence of resolution feels true rather than evasive. The reader does not close the book frustrated, even though it is finished.
This requires structural understanding of which threads the story has pursued and which it has only gestured toward, which mysteries the narrative has made load-bearing and which can remain open. The writer must ask whether what they have written corresponds to what people do, feel, and fail to resolve, rather than what the narrative logic seems to demand. And it calls for the confidence to hand the reader something incomplete, trusting them to hold it.
A language model has none of this. Confidence of this kind is not statistical; it is an act of faith in the reader. It belongs to a writer who has finished books that did not explain everything and in that absence found more honesty than explanation would have.
The model closes every door. The writer knows which ones to leave open.
Rio de Janeiro, the iv day of July, MMXXVI.
