Notes · On writing with AI

A humanizer hides the machine. Your reader was never looking for the machine.

July 2026 · 5 min read · by Leigh Daniel

There is a whole category of tools now that take finished AI text and rewrite it until a detector scores it as human. Monthly subscriptions, per-word pricing, user counts in the millions. The pitch is concealment: nobody will know a machine was involved. And the entire category is aimed at a problem you do not have.

Think about who actually reads your work. Not in the abstract. The specific people: your boss, the VP two levels up, the client who signs the renewal. None of them runs your memo through detection software. They read the first two sentences, and somewhere in those two sentences they form an opinion, not of the memo, of you. That opinion is the only detector in the building, and no rewriting tool can reach it.

Because here's what a humanizer actually produces. It takes text that reads like a machine and sands it until it reads like a person. Any person. A committee of no one. The vocabulary gets shuffled, the rhythm gets roughed up just enough to fool the classifier, and what comes out the other side is writing with a pulse and no fingerprints. It scores as human. It does not score as you, and yours is the name at the top.

Text that reads as “a person” is not the same as text that reads as you. Your name is on it either way.

There's also the quieter cost, the one the pricing page doesn't mention. The moment your plan becomes hiding that AI was involved, you've conceded the point. You've accepted that the writing isn't yours and paid a second subscription to keep anyone from noticing. Two tools deep, twenty dollars a month, and the output sounds like nobody. The people who get caught in this loop aren't lazy. They're solving the wrong problem with real diligence.

The actual problem sits one step earlier, before the draft exists. AI output sounds generic because the machine has no idea how you write. It has never seen your sentences. So it produces the average of everyone, and then you either ship that average or spend twenty minutes hand-editing it back toward yourself, every draft, forever.

The fix at that step is not concealment. It's specification. Capture how you actually write: the rhythm, the words you reach for, the shapes your sentences make, and hand that to the machine before it writes a line. Then the draft starts in your voice instead of being laundered into a stranger's afterward. That's the whole design of the Voice Kit, and if you want the two approaches laid side by side, feature for feature, the comparison page does exactly that.

Concealment is a subscription. Sounding like yourself is a one-time fix.

See the full side-by-side at the Voice Kit vs. a humanizer, or check what your current drafts are telegraphing with the Slop Check, free, in your browser.