Every: Editing AI writing

Every is a media company that has written publicly and in detail about how it uses AI across editorial work. This piece from the Context Window newsletter focuses on the editorial side—specifically, what changed when Every built a structured style guide and fed it into Claude so that writers could check their own drafts before reaching an editor.

The key figure in the piece is Kate Lee, Every’s editor-in-chief. Lee describes the 400-rule style guide as something that grew from necessity: as AI tools made draft generation faster, the number of drafts arriving at the editorial stage increased, and the errors that came with them were often stylistic rather than structural. A guide that writers could run against a draft before submission reduced the rate of those errors and shifted editorial attention toward the work that required genuine human judgment—argument quality, narrative coherence, perspective.

The piece also makes a point about AI adoption that cuts against the productivity-first framing common in discussions of AI in the workplace. Lee’s own breakthrough moment with AI tools came not from writing assistance but from using an AI agent to manage hiring logistics in Notion. The tools that stuck were the ones that solved specific, painful problems rather than the ones that promised general efficiency.

Contributor Eleanor Warnock adds a perspective on AI-generated text that is worth noting separately. Her argument is that raw AI output—even with good guidance—often lacks substance, and that the quality ceiling is set early, by how much deliberate preparation goes into the system before it generates anything. Studying examples of writing you want to emulate, building those into your prompts, and treating the AI as a partner that needs context rather than a tool that self-orients all belong to the preparation phase, not the editing phase.

Who this is useful for. Editors and editorial directors at publications experimenting with AI-assisted drafting, content leads at companies trying to maintain voice consistency at higher output volumes, and writers who want a clearer picture of what AI integration at a successful media company actually looks like in practice.