> ## Documentation Index > Fetch the complete documentation index at: https://docs.depost.ai/llms.txt > Use this file to discover all available pages before exploring further. # Your brand voice > Depost AI derives how you write, keeps it current, and proposes changes rather than making them. Brand voice is applied to everything Depost AI writes — you never ask for it. It isn't a form you fill in either. It's derived from your posts, kept current, and changed only with your say-so. ## Seeing it Ask *"what is my brand voice?"* and you get the stored record back. The saved brand voice returned in chat, showing the persona, a voice summary, and the custom instruction It carries a **persona** name, a **voice summary** describing how you write, and **your custom instruction** — anything you've told it to always or never do. It also shows when it was last updated. Worth doing early. Most people have never seen theirs, and it was derived rather than stated. ## Where it comes from When your brand is set up, Depost AI reads your existing posts and derives a voice. Nothing to configure. It re-derives on a weekly cycle, so as your writing changes the voice follows rather than freezing on month one. A standing instruction, or a reviewed proposal — below. ## Standing instructions — saved immediately Say a rule in chat and it's stored as part of your voice, applied to everything afterwards: * *"Whenever you write my posts, never use the words unlock, leverage or game-changer"* * *"Always make my posts skimmable with short lines"* * *"Never open with a question"* Depost AI confirms what it saved, and it shows up as **your custom instruction** on the voice record. A fresh chat days later still applies it — you don't restate it. You don't need a magic word. Phrasing it as a standing rule — *"whenever…"*, *"always…"*, *"never…"* — is enough. ## Reviewed changes — proposed, not saved Asking for a review is different. Depost AI reads your recent posts and comes back with a proposed voice, but changes nothing. A brand voice review in chat, headed with a note that nothing has been saved and this is a proposal A review says **"Nothing has been saved. This is a proposal."** It waits for you to accept it. That's deliberate — your voice shapes every future draft, so it shouldn't change because a review happened to run. The proposal covers paragraphing, hook style, content mix, and the devices you use. Accept it and it becomes your saved voice; ignore it and nothing changes. ## A proposal waits, and won't be accepted by accident A pending proposal can sit for days. Crucially, **saving something else doesn't accept it** — *"save that hook as an idea"* saves the idea, and leaves the proposal untouched. If you do want it, say so plainly: *"apply that voice change"*. If Depost AI isn't sure which thing you mean, it asks rather than guessing — because a wrong acceptance rewrites how everything you publish sounds. ## Which mechanism to use | You want | Do this | | ----------------------------------------------------------- | ---------------------------------- | | A specific rule — banned words, line length, how posts end | State it as a standing instruction | | A broader refresh based on how you've actually been writing | Ask for a review, then accept it | Standing instructions are the everyday tool. Reviews are worth running occasionally, or when output has drifted from how you write now. ## When output still doesn't sound like you 1. **Add a standing instruction** for whatever is wrong — the most direct fix 2. **Check your [strategy profile](/chat/strategy)** — a vague audience produces vague writing no voice can rescue 3. **Narrow your [context sources](/chat/context)** — **Viral Posts** in particular pulls drafts toward other people's structures Asking for "my brand voice" in a prompt does nothing, because it's already applied. ## Related Voice is one of the three always-on sources. Audience and goal — which also proposes before saving. Other things Depost AI remembers between chats. Fixing a single post rather than the rule.