Voice Preservation Checklist
About this resource
There is a version of AI-assisted writing that saves time and slowly erases the writer. The output is competent, grammatical, and could have come from anyone, which for a leader communicating with their own team is a real cost rather than a stylistic quibble. People follow a voice they recognize.
This checklist covers keeping yours. It has three parts: what to give the model up front so the draft starts closer to you, what to check before anything goes out, and the categories of writing that should not be delegated at all regardless of time pressure. It is the constructive counterpart to The Tics That Give ChatGPT Away: that handout tells you what to suppress, this one tells you what to preserve.
Supply your voice before drafting, check it before sending, and know which writing should never be delegated. The third part is the one people skip.
How to use it
Build the voice brief once and save it in your assistant’s custom instructions. Two paragraphs of your own writing plus three lines about what you avoid will do more than any per-message instruction, because the per-message version gets skipped exactly when you are busy. Then keep the do-not-delegate list somewhere you will actually see it, because that decision gets made under time pressure or not at all.
Inside the download
What to give the model so the first draft starts closer to you.
Why two paragraphs of your actual writing beats any description of your tone.
Protecting the phrases and rhythms that make writing recognizably yours.
The fastest check available, and the one people skip.
Common questions
Give the model samples rather than descriptions. Two or three paragraphs of your actual writing communicates your voice far more precisely than any adjective, and saving them in custom instructions means you do it once. Then read the result aloud before sending; if it does not sound like you speaking, it is not finished.
Anything where the fact that you wrote it is the point. Apologies, condolences, personal praise, difficult feedback to someone you manage, and anything acknowledging your own mistake. In each case delegation removes the thing that gave the message its value, and people can usually tell.
It defaults to a recognizable register because it was trained to produce fluent, professional prose. That default is avoidable, but only deliberately: supply your own samples, name what you avoid, and edit the result rather than sending the first draft.
Free. No email list, no follow-up sequence.
More from the resource library, chosen because they pair well with this one.
Seven habits that make AI writing obvious, and one copy-and-paste instruction that switches all of them off.
Design a reusable AI persona so a whole team gets consistent output without everyone learning to prompt.
What to say when your team asks the hard questions. Six conversations, with the language that works and the language that backfires.
Tell us a little about yourself and this document is yours. We don't have an email list. We collect this so we know who's finding our resources useful, and that's it.
You're downloading
this document
Your information is kept confidential and never shared with anyone.
