“I teach a particular way of writing a LinkedIn article — clean transitions, well-worked-out logic. Now an article I follow says that exact structure reads as AI-generated, because it’s too perfect. Humans don’t write that cleanly, so people assume it must be AI. The recommendation: humanize it — bring in things an AI couldn’t possibly know.” — Don Back
Session context: 2026-06-18_Mastermind — Don Back surfaced this while modifying one of his short courses, having just read that his own taught structure had become a detection signal.
Core Idea
There’s a cruel inversion happening in content: the polished structure that used to signal competence now signals automation. Clean transitions, airtight logic, perfectly balanced paragraphs — the very craft markers educators (Don included) taught as best practice — are now what makes readers think “an AI wrote this.” Perfection became suspicious. The structure didn’t get worse; the context changed underneath it.
This matters because the cost isn’t aesthetic, it’s trust. When a reader pattern-matches your post to “AI slop,” they discount it before they engage with the idea — and for a knowledge entrepreneur, perceived authenticity is the product. The fix isn’t to write worse. It’s to add the signals a model structurally cannot fake: lived specifics, things that actually happened, a conversational register, the small imperfections and asides that mark a human present in the text. The mechanism is exactly the one behind voice-of-customer work — concrete, particular, first-hand language is unforgeable; generic competence is now cheap and therefore cheapening.
The deeper, slightly uncomfortable lesson for anyone who teaches frameworks: a best practice is contingent on its environment. The moment a structure becomes the default output of the tools everyone uses, teaching it as a differentiator stops working and starts back-firing. Best practices now have a half-life, and authenticity markers are the new differentiator precisely because they don’t automate.
Practical Application
Run the “AI tell” pass on your next post before publishing. Ask: what here could only have come from a machine that’s read everything and lived nothing? Then inject what it can’t: one specific thing that actually happened to you (a date, a name, a number, a failure), a conversational aside or incomplete thought, an opinion stated without hedging. Don’t sand the edges off — leave the human texture in. If you teach writing, teach the authenticity layer over the structure layer, and tell students plainly that the clean-structure era is a detection liability now.
Related Insights
- Insight - AI as Ghostwriter, You as Editor-in-Chief — the human’s job moves from producing structure to supplying the unfakeable specifics and judgment.
- Insight - The AI Authorship Identity Crisis — When Creators Can’t Tell Where They End and the Tool Begins — the same fault line from the creator’s side; this is the reader’s side.
- Insight - Authentic AI Voice Is Built on Lived Experience, Not Style Prompts — lived experience is the unforgeable signal; this is its inverse, the absence of it reading as a machine tell.
Evolution Across Sessions
First articulation of structure-as-AI-tell in the vault. Where prior insights framed authenticity as a creative-identity question, this names a concrete, current market mechanic: a craft norm flipped from asset to liability because the tools made it the default. Establishes the baseline — authenticity markers, not structural polish, are now the differentiator — for a theme likely to recur as detection norms keep shifting.