The single most commonly identified AI writing tell. Negative parallelism with an em-dash reframe. Humans occasionally use antithesis; AI uses it in back-to-back sentences. Wikipedia editors have seen LLM-generated text that uses it in literally every sentence.
The Pattern
"It's not X -- it's Y" is antithesis. Humans have used it for thousands of years. Nothing wrong with the device itself.
What gives it away is frequency. AI will use this construction multiple times in a single paragraph. Sometimes in every sentence. Humans deploy it for deliberate emphasis -- once, maybe twice in an essay. AI treats it as a default sentence template.
Blake Stockton called negation "AI's single most beloved technique" and documented its dominance across ChatGPT outputs.
Wikipedia editors flag it as a primary detection signal. They've seen articles where every single sentence runs through some variant of "not X, but Y."
RLHF training rewards the structure because the negation-reframe feels insightful and corrective. Raters like it. So the model learns that negation-then-correction sounds smart and keeps reaching for it.
Variants: "It's not just X -- it's Y", "X isn't about Y -- it's about Z", "The problem isn't X. It's Y.", "Forget X. Think Y."
Examples
The Research
Blake Stockton's "Don't Write Like AI" series called negation the single most prominent AI writing technique. ChatGPT defaults to negation-reframe whenever it wants a point to feel insightful. It's the rhetorical equivalent of a crutch.
Wikipedia's "Signs of AI writing" guide lists this pattern front and center. Editors report seeing it in a high percentage of AI-generated article submissions. Antithesis in human writing is a choice. In AI writing, it's a tic.
Colin Gorrie nailed the diagnosis: AI knows classical rhetorical techniques -- antithesis, parallelism, chiasmus -- but has no sense of when to use them. "What the LLM lacks is not technical ability, but taste -- the taste to know when to deploy these techniques." It's a musician who knows every chord but plays all of them in every song.
tropes.fyi catalogs it under "Negative Parallelism" as a core sentence-level pattern.
Caught in the Wild
WikiProject AI Cleanup uses "not X, it's Y" as a primary detection signal. Articles with the construction in every paragraph get flagged and reverted on sight. The pattern was cited as evidence when Wikipedia voted to ban AI-generated article content in March 2026 (44-2).
TechCrunch →Negation-reframe became the signature move of AI-generated LinkedIn. Originality.ai found 54% of long-form LinkedIn posts were AI-generated, and the "not X -- it's Y" structure showed up at far higher rates in those posts than in human-written ones.
Originality.ai →Writing coaches and content guides now explicitly ban "It's not X -- it's Y." Jodie Cook's widely-shared "ChatGPT ban list" files it under "False Directness." When your rhetorical device has its own entry on a ban list, it's been overplayed.
Jodie Cook →Sources