AI for Business
Story & Draft Editor (an edit, not a rewrite)
Paste your own draft and get an editor’s read: what is working and where, what is not, line-level suggestions as your line and a proposed change side by side, pacing notes, and the three things to fix first. It does not rewrite you, and it will not hand back a smoothed-out version of your voice.
This is a cloud tool. Your draft is sent to the model to be read. It is not stored by us and Anthropic’s API terms do not use it to train models — but it is unpublished work leaving your device, and that is your decision to make each time. Nothing is stored by us; the call is counted against your month’s allowance — 10 free, 300 on Pro. How the AI tools handle data.
Tips
- It is an edit, not a rewrite. Every suggestion is your own line quoted back with the smallest change that fixes one thing — never a smoothed-over version of your paragraph, and never a new draft you have to argue with.
- The "leave these alone" list is the one to read first. It names the things that look like mistakes and are not — the fragment, the repetition, the unfashionable word — so that nothing in the notes talks you out of your own voice.
- Tell it what is deliberate in the "anything I should know" box. An editor who does not know the ending is meant to be flat will spend three notes trying to fix it.
- Change the stage and the notes change with it. A first draft gets structural notes; a nearly final one gets the lines, because by then the structure is not the question.
- The three things to fix first are ranked on purpose. Everything else is a list; that is a plan.
Frequently asked questions
Will it rewrite my work?
No, and it is told so in the strongest terms in its instructions: it is an editor, not a ghostwriter, it never returns a rewritten version of the piece, and its line suggestions are the smallest change that fixes one specific problem. If a suggestion ever reads as "here is how I would have written it", that is the suggestion to ignore — and it is the failure the tool is built to avoid.
Does it understand what I am doing stylistically?
Partly. It is good at noticing rhythm, repetition and register, and it is told to protect them rather than normalise them. It is not good at knowing what your book is for. Use the "anything I should know" box and the questions it asks you at the end; those are the places where it admits what it cannot see.
Is my unpublished work safe?
The draft goes to the model through our gateway to be read, and comes straight back. We store neither the draft nor the notes, and Anthropic’s API terms do not use API content to train models. That is the honest position: it left your device, it was read, nothing was kept. Whether that is acceptable for a manuscript under submission is a judgement only you can make.
How long a piece can it read?
About 8,000 words a run, which is a chapter or a long story. A whole novel is not a job for one pass; edit it a chapter at a time, and the notes stay specific instead of becoming generalities.