Briefing: You Write Like a Machine. That's Good.

Published: May 5, 2026 | Source: ejsays.com | Author: E. J. Original article: https://posts.ejsays.com/you-write-like-a-machine-thats-good/

Core claim: AI detection systems don't detect AI. They detect pattern — specifically, coherence, logical structure, and lexical precision. These are the exact features LLMs were trained to replicate from human writing. The classification boundary between human and AI authorship is indeterminate by design. The merchant who sells both the sharpest spear and the strongest shield already knows his answer. He rebranded as a SaaS company.

The Tuesday morning incident:

• 1,110-word article submitted by a student: 31% AI likelihood
• Editor (the author) spent ~10 minutes fixing logic sequences only — no vocabulary changes, no grammar rewrites
• Article resubmitted: 67% AI likelihood
• Churchill's writing fed into the same system: flagged as AI-generated

The Han Fei frame (c. 250 BCE): A merchant in the state of Chu sold a spear that could pierce any shield, and a shield that could block any spear. A bystander asked: what happens when your spear meets your shield? The merchant had no answer. He still doesn't. He rebranded as a SaaS company selling humanization services.

Why the business model requires high scores: A detector that cleared most articles would have no product to sell. Detection is often free. Humanization is always paid. The threshold is set where it is to sell humanize — not to detect accurately.

The full loop:

The structural contradiction: LLMs were trained on human writing via RLHF — optimized to approximate human linguistic expression. Detection systems flag logical coherence, structural consistency, and lexical precision as AI indicators. These are precisely the features LLMs learned from human sources. The optimization target of the generative system and the detection criteria of the evaluative system converge on identical textual properties. In many cases, the model writing the draft, the model judging the draft, and the model humanizing the draft are architecturally identical. The user is charged three times for the same spear, shield, and upgrade.

The Churchill result: Winston Churchill — the man who held the English language together during the worst years of the twentieth century — does not write human enough to pass. High AI score = you write with structure and intention. The machine spent billions learning to do what Churchill did naturally.

Who has the right to determine whether a person's style is too human or too AI? The author's answer: not the company that also sells humanization to you.

To teachers and editors: Read the work. A high AI score tells you the writing is coherent. It tells you the argument flows. It tells you nothing about whether a human thought the thoughts. If you trust a percentage more than your own reading, you have already handed your judgment to a machine.

To creators: Stop apologizing. Don't dumb it down. Don't hollow out the logic to satisfy an algorithm. Wear the score. It means you still know how to think.

AI Detection Business Model