There is an article making the rounds right now about how to spot AI writing.
The author is a professional editor. Smart person. Good eye. She identifies the tells. The em dash. The word “delves.” And the one that set her off — the sentence structure that goes “It’s not X. It’s Y.”
She’s right about all of it.
But the fix she’s pointing toward is a scrub job. Pull the tells. Sand down the surface. Make it passably human.
That’s not a fix. That’s cosmetic work on a structural problem.
Here’s what the scrub job misses.
The AI writing problem isn’t the words. It’s the cadence. Every AI-generated paragraph wants to build the same way. Setup. Pivot. Payoff. Setup. Pivot. Payoff. The reader doesn’t consciously notice it. They feel it. It reads like someone who learned to write by studying effective writing rather than by having something to say.
Pull the em dash. The cadence remains.
Pull “delves.” The cadence remains.
Pull “It’s not X. It’s Y.” The cadence remains.
The surface is clean. The engine underneath is still running the same loop.
The article cites a writer named Sam Kriss who goes deeper than most. He’s not counting word frequencies or flagging punctuation. He’s looking at the larger patterns. The mixed metaphors. The rule of three that shows up in almost every AI paragraph. The phrases borrowed from one culture dropped into another where they land wrong.
His diagnosis is accurate. AI writing lacks direct experience of anything. It has never been cold. Never been embarrassed. Never sat in a waiting room with bad news coming. So it reaches for pattern instead of memory. It tells you something is significant rather than showing you the moment that made it significant.
That gap is what readers feel. Even when they can’t name it.
The editor in the article scrubbed her client’s document. Removed the obvious tells. Checked for hallucinations. Did everything right by the standard checklist. The document still sounded like a machine wrote it.
Because it did.
And no checklist fixes that.
Here is the structural problem underneath the surface problem.
AI is trained on enormous amounts of human writing. It learned what effective writing looks like from the outside. It learned the shape of a good argument. The shape of a compelling story. The shape of a sentence that lands.
What it did not learn is what it feels like to have something to say and not enough words to say it right. What it feels like to write the wrong sentence and know it’s wrong before you finish it. What it feels like to sit with a thought until the right line arrives.
Those experiences produce a different kind of writing. Not because of the words they choose. Because of the pressure behind the words.
AI writing has no pressure behind it. It has pattern. Pattern is fast. Pattern is consistent. Pattern produces something that looks like writing from a distance.
Up close it feels like a machine that learned to impersonate someone who has something to say.
The scrub job makes it look less like that. It does not make it be less like that.
The Baseline voice standard addresses the engine. Not the surface.
Voice-first means the line comes from a human voice and the writing builds outward from it. Not argument first with voice filled in afterward. Not a framework assembled and then softened to sound personal. A live line. A real thought. Building outward from there.
That changes the cadence because it changes whose rhythm is running.
When a human voice drives the line, the AI is working against its own default. The pattern-matching reflex has something to push against. The setup-pivot-payoff loop gets interrupted because the opening doesn’t set it up. The structure has to follow the voice rather than the voice being fitted to the structure.
That is the only correction that works at the level where the problem actually lives.
Not a style guide. Not a banned word list. Not a scrub job run after the machine has already done its work.
A different process from the start.
The honest thing to say about AI writing tools — and this applies to every platform, every model, every workflow that hands the first draft to a machine — is that the output carries the machine’s signature. Always. The question is how deep that signature goes and how much of the writing process was human before the machine touched it.
A human who writes their own first draft and asks AI to improve the grammar has a document with a light machine signature. A human who gives AI a topic and publishes what comes back has a document that is, underneath whatever scrubbing followed, a machine document.
Most of what is being published right now under human names sits somewhere between those two points.
The readers know. Not always consciously. Not always in words. But the feeling is there. The cadence gives it away even when every individual tell has been removed. Because the cadence is not a feature of the words. It is a feature of the process that produced them.
Change the process. Change the cadence.
That is what the Baseline voice standard is built to do.
Not make AI writing sound more human by removing the parts that sound inhuman. Make the writing human by keeping the human in the driver’s seat from the first line to the last.
The scrub job works on the output.
The voice standard works on the origin.
Those are not the same thing. They never were. And every reader who has ever put down a piece of content because something felt off — without being able to say what — already knows the difference.
They just didn’t have the words for it until now.
“The Faust Baseline Codex 3.5”
Author of the category ”AI Baseline Governance”
Post Library – Intelligent People Assume Nothing
“Your Pathway to a Better AI Experence”
Purchasing Page – Intelligent People Assume Nothing
Unauthorized commercial use prohibited. © 2026 The Faust Baseline LLC
