An AI copy editor in your notes (that actually works)

AI can function as a practical copy editor inside a notes app—without rewriting someone’s draft from scratch—so writers can keep their voice while fixing spelling, grammar, punctuation, and paragraph structure. The workflow demonstrated uses Reflect, which includes GPT-4, to select existing text, run an “AI editor” action, and receive an edited version that preserves the original writing while tightening errors and formatting.

The core value is control. Instead of generating a new article, the process starts with a draft that already exists (even if it’s “pretty poor”), then applies editing layers. In the first pass, Reflect’s built-in copy editor focuses on mechanical improvements: spelling mistakes, grammar issues, punctuation, and appropriate paragraph sizing. The output is essentially the same writing, but cleaner—enough that a user could stop after the first edit if they only need polish.

From there, the transcript shifts to customization: building a personal copy editor prompt that goes beyond basic proofreading. The example uses a writing-tactics list pulled from a book called “The Subtle Art of Writing” (titled in the transcript as “the symptoms Style,” described as “a guide to writing in the 21st century”). Highlights from the book are converted into a set of writing guidelines, then pasted into a cloned copy-editor prompt. The result is an editor that not only corrects errors but also rephrases sentences, changes word choice, and adjusts flow—adding examples and restructuring phrasing in a way that still aims to keep the user’s underlying intent.

A key operational detail is how revisions are managed. After running the customized editor, the writer doesn’t automatically replace the draft. They review the edited version, rerun if the opening doesn’t work, and then selectively merge: keep the parts they like, rewrite the parts they don’t, and iterate a few times. The transcript argues this can feel faster than manual editing because the AI handles the heavy lifting, while the human focuses on judgment—what fits the desired style and what doesn’t.

The customization possibilities extend to different writing contexts. The transcript suggests tailoring separate “copy editors” for different tones and formats—business writing, blogging, social media, or even tweet-length constraints—by treating the prompt like a job description for the kind of editor the user wants. The overall takeaway is that AI editing in notes can be both lightweight (proofreading) and deeply configurable (style-driven rewriting), as long as the workflow preserves the writer’s draft and uses iterative selection rather than wholesale replacement.