The moment your AI passes for you.

The journalist held up two paragraphs. Both were written in the same voice — the rhythm, the asides, the slightly impatient turn at the end of a sentence. One was written by the consultant herself. The other was generated by her Brain.

She read both. She paused. She said, “Honestly, I couldn't tell either.”

This is the moment we call the Turing-pass. Not the Turing test in the formal sense — that one is about whether a machine convinces you it is human. The Turing-pass is more specific. It is about whether an AI sounds like a particular human, indistinguishably enough that the human herself can't tell anymore.

For an expert, this is the line between a tool and a system. ChatGPT can sound smart. Claude can sound articulate. Gemini and Perplexity can marshal every fact on the internet. None of them can sound like you — because none of them has been trained on you. They have been trained on the internet. Your specific cadence, your hedges, your jokes, the way you build to a recommendation rather than leading with it — none of that is in their weights.

To pass for you, the system needs your archive. Years of writing. Voice memos describing how you actually think. Briefs you've sent. Posts you've discarded. The forbidden phrasings (“best practices” — Ghalia bans it). The signature turns (“trois choses, dans l'ordre”).

When the archive is dense enough, and the calibration loop tight enough, the output stops sounding like AI and starts sounding like a stranger reading your own writing back to you.

Most expert practices have an archive. Few have ever turned it into an asset.