Draft
I ran my latest essay through Substack's AI scanner. 100% AI. It's right, and it still missed the part that mattered.
Last week Bloomberg Businessweek wrote about how I write and asked the same question: "Does most of this writing read like AI? You bet."
For years the comfortable answer was that detectors don't work. That answer expired this year. They're good now, and what they measure is real: how much of the wording came from a machine. My process starts with twenty minutes talking into a recorder, arguing the piece out before any model is in the room. The scanner only sees what came after.
At work, and in a post like this one, writing is how an idea gets communicated, and the idea is what gets judged. Score the wording and you miss the forest for the trees. The forest is the thinking behind the writing and someone willing to own it. The test I'd use is older than any of this: if I ask you ten questions about what you wrote and you can't answer them, that's the problem.
I also wrote about why I'd rather own a piece than stamp a disclaimer on it, and how building my own writing system changed how we think about voice at Typeface.
Full piece is on Substack. Link in comments.