A short reflection on style, substance, and (perhaps) integrity
Using AI to assist in my writing has been unexpectedly transformative.
For years, I wrote alone—drafting, revising, structuring, revisiting old pieces, tightening arguments, clarifying evidence. Then I began asking a chatbot to rewrite sections in my style, to try a second draft with my cadence, or to expand a paragraph into an argument worth making. The shift has been to greatly speed the writing process: not because the machine has replaced the work, but because it helps to get to much of the real work faster.
What I have learned, though, is that the machine’s eagerness to please is both a strength and a liability.
Give it a paragraph of my writing and say “rewrite this in my style,” and it does so with enthusiasm—sometimes too much enthusiasm. It smooths language, but also softens edges. It amplifies insight, but sometimes with a kind of rhetorical helium that lifts the argument a little too high. It loves a flourish. It dislikes ambiguity. It adds connective sentences where I would have left space. It turns a sharp line into something slightly sentimental unless I rein it back.
Most importantly, when something in my text is ambiguous, the machine often refuses to leave well enough alone. It fills gaps. It embellishes caveats. It asserts where I intended to speculate. It turns the grey zones of policy analysis into black-and-white statements, not maliciously perhaps, but because its training pushes it toward completion rather than restraint.
So the work—the real work—is in the original ideas the reframing and in revision.
I edit out the “clunkers”: the platitudes, the overly rounded sentences, the attempts to make the argument more uplifting or stronger than I intended. I strip away the false certainty. I restore ambiguity where ambiguity belongs. I take what is useful—structure, phrasing, rhythm—and remove what is not. When it loses my voice, I put it back in.
This is not fundamentally different from how writers have always worked.
We lean on:
- outlines,
- memos,
- prior drafts,
- editors,
- colleagues,
- and sometimes our own past writing.
The machine can be another tool in that lineage—one that produces a very fast first draft.
Some observers worry that this is plagiarism. It is not.
Plagiarism requires passing off someone else’s work as one’s own without attribution.
But in my case and that of many others, the “someone else” is not producing original ideas, arguments, or research. The machine is rewriting my thoughts, using my words, guided by pieces I have previously written, often with explicit instruction to “use my own style.”
When I ask it to rewrite something using my cadence, I am not stealing—I believe I am borrowing from myself. I still worry about it but I am worrying about it less; maybe a fool’s paradise? Maybe not.
This appears to be no more plagiarism than asking an editor to re-structure a paragraph, or using a thesaurus, or learning from the rhythm of one’s earlier publications. If anything, the risk is the opposite: the machine becomes too much like me, echoing patterns I recognize, borrowing my structures a little too faithfully.
The ethical use of AI in writing lies not in avoiding the tool, but in refusing to let it replace judgement.
I retain ownership of:
- the ideas,
- the framing,
- the research,
- the evidence, and
- the final editorial decisions.
The machine accelerates drafting. I do the thinking.
What AI cannot do—and what it must never be trusted to do—is evaluate truth claims or fill in missing evidence. It has no concept of “don’t know.” It assumes completion is better than uncertainty. But in policy work, uncertainty is essential. It marks the boundaries of evidence. It tells readers where caution is required. It is the intellectual honesty that should prevent error.
So I treat the machine not as an authority but as a writing partner who is brilliant at speed and mediocre at judgement. I accept its assistance and reject its overreach. I let it generate a draft but insist that the final piece passes through a human filter—my own—where evidence is checked, arguments are sharpened, ambiguities are restored, and the tone is brought back into alignment with what I actually believe.
Using AI to accelerate writing is neither improper nor plagiarism when the ideas are mine, the structure is mine, the judgement is mine, and the drafting process is transparent to me. It is simply a new kind of pencil—faster, yes, but it still appears to be incapable of replacing the hand that holds it.