Every: My editor caught me sounding like AI. Now AI catches me first.
Katie Parrott, writing in Every’s Working Overtime newsletter on June 8, 2026, describes what happened when her editor started flagging patterns in her drafts that read as AI-generated: symmetrical sentence structures, rhetorical sets of three, and other tells that crept in as she used AI more in her writing process. Rather than abandoning AI assistance or leaving the flagging to her editor, she built a system for catching those patterns herself before the draft left her hands.
What the article is about
The article describes a practical workflow using what Parrott calls “skills” — reusable AI reviewer personas, each with a distinct mandate. She built four:
Sorkin maintains narrative momentum and flags passages where unnecessary detail slows the piece down.
Mom reads for general audience comprehension, identifying sections that assume too much prior knowledge.
Asshole attacks the weakest version of the argument, surfacing counterarguments the writer has not addressed.
/guardrails catches mechanical AI tells: symmetrical sentence structures, sets of three, and other syntactic patterns that accumulate when AI has touched a draft.
These reviewers operate at different stages: an outline pass where the argument gets pressure-tested, a section-by-section pass for prose cleanup, a full draft pass for big-picture coherence, and a final comprehensive checklist.
Key takeaways
The most useful insight is structural: documenting your editorial standards explicitly, in a form that an AI can act on, makes those standards clearer to you as well as enforceable by the AI. Most writers carry their standards implicitly, which makes them inconsistent. Writing them down for an AI reviewer creates a record that can be revised as the writer’s voice develops.
The second insight is about reviewer diversity. No single AI reviewer catches everything, because reviewers optimized for different concerns surface different problems. A reviewer focused on argument strength will not notice prose rhythm; a reviewer focused on AI tells will not notice whether the argument holds.
Parrott provides an open-source toolkit that readers can adapt to their own standards.
Who it is useful for
Writers who use AI in their drafting process and want a systematic method for maintaining editorial quality and voice consistency. Also useful for editors building review workflows for teams that produce AI-assisted content, where the goal is consistent standards rather than per-piece intervention.