Capture how they sound — in 3 hours

What it is: The top-of-doc quick reference — 3-5 adjectives that describe the voice, plus a "sounds like / never sounds like" line. Scannable in 30 seconds.

Why it earns premium: It's the part a new team member, another VA, or an AI can absorb instantly. The whole doc in one glance.

What to include:

🤖 AI prompt:

"Here are samples of my client's own writing: [PASTE 3-5 SAMPLES]. Analyze the voice and give me: (1) 3-5 adjectives that describe how they sound, (2) a single 'sounds like ___, never like ___' line, and (3) one sentence naming who this voice is speaking to. Base everything on the samples — don't invent traits that aren't there."

What it is: Where the client sits on key spectrums — formal↔casual, serious↔playful, concise↔detailed, warm↔direct — with notes on how tone shifts by channel.

Why it earns premium: Tone is what varies by context. Capturing it prevents the "too stiff for our Instagram / too loose for our proposal" rework.

What to include:

🤖 AI prompt:

"Based on these client samples [PASTE SAMPLES], place the voice on these spectrums and explain each in one line: formal↔casual, serious↔playful, concise↔detailed, warm↔direct. Note any place the tone clearly shifts by channel (e.g., email vs. social). Use only what the samples show."

What it is: The fingerprint — signature words and phrases the client uses, words they'd never use, how they refer to their product and customers, plus emoji and punctuation rules.

Why it earns premium: Specific words are what make content recognizable. This is the section that kills the "this doesn't sound like us" feeling fastest.

What to include:

🤖 AI prompt:

"From these client samples [PASTE SAMPLES], extract: (1) signature words and phrases they repeat, (2) words or phrases that would feel off-brand for them, (3) how they refer to their product/service and their customers, (4) their emoji and punctuation habits. List them plainly. Flag anything distinctive."

What it is: 3-5 real samples of the client's best writing, annotated for WHY each one sounds right. Optionally, a before/after rewrite showing off-voice → on-voice.

Why it earns premium: Examples train both humans and AI faster than rules ever could. This is the highest-value section of the doc.

What to include:

🤖 AI prompt:

"Here are strong examples of my client's voice [PASTE SAMPLES]. For each, write a one-line annotation explaining what specifically makes it sound like them (word choice, rhythm, structure, attitude). Then take this off-voice draft [PASTE A GENERIC VERSION] and rewrite it to match their voice, so I can show a before/after."

What it is: A ready-to-paste prompt block that encodes the whole doc, so anyone — the client, you, another VA — can feed it to AI and get on-voice output instantly. It turns a static document into a working tool.

Why it earns premium: This is what makes the doc alive. A style guide sits in a drawer; a Voice Prompt gets used every day.

What to include:

🤖 AI prompt:

"Using this Brand Voice Doc [PASTE COMPONENTS 1-4], write me a single reusable AI prompt I can paste at the start of any content request. It should tell the AI how this brand sounds (tone, vocabulary, do's and don'ts), include 1-2 short examples, and instruct it to match this voice in everything it writes. Make it copy-paste ready."

Most VAs should start at Standard.

You're CAPTURING a voice, not INVENTING one. That distinction matters more than any other rule this week.

If you're new to voice work or building this for the first time: start with a sample. Build the doc for a public brand you admire. Show the result. Then offer the same service to a real client — discounted first time, full price the second.

If you already work with the client: propose the doc as a one-time deliverable that makes all future content faster. Frame it as "I want to capture how you sound so anything I create — or anyone else creates — stays on-voice."

Either way, the doc ends with a client-approval step. "Here's how I've captured your voice — anything feel off?" That step isn't a weakness. It's the move that makes the doc theirs. And it makes you the person who checks.

This is what makes the time math real:

Step 1 — Gather raw material: 5-10 pieces of the client's own writing (emails, posts, website copy, past messages). This is the whole secret. You capture, you don't invent.

Step 2 — Run Components 1-4 through the AI prompts, one at a time, feeding the same samples.

Step 3 — Curate: keep the quirks, cut anything generic, fix anything AI got wrong. The weird specific phrases are the gold. Protect them.

Step 4 — Compile Component 5 (the Voice Prompt).

Step 5 — End with the client-approval step. Send the doc. Ask what feels off. Adjust.

Master template build with AI: ~3 hours.
Per-client customization after that: ~30 minutes.

_{If you have active clients now: Pick a client who's ever said "not quite us" — or one about to add more help. Propose the Voice Doc as a one-time deliverable.}

_{If you don't have clients yet: Build a SAMPLE Voice Doc analyzing a public brand you admire. Goes straight in your portfolio. Answers the silent question every first client asks: "can this person sound like us?"}

The Voice Doc workflow above is AI-agnostic — works with ChatGPT, Claude, Gemini, anything.

If you want the Claude-specific deep build — Projects for client workspaces, Skills that persist your voice rules across chats, multi-tool orchestration — that's where the Claude for VAs 30-Day Bootcamp picks up. By Day 30: a portfolio of client-ready deliverables, your packaged AI service, and UpSkill AI certification. Founding cohort opening soon.

Join the Bootcamp waitlist →

Tomorrow: "Who am I to define someone's brand voice?" — and what to do when you're scared you'll get it wrong.

P.S. — If you're a VA who wants to stop worrying about AI and start earning more because of it, 200+ Filipinas inside UpSkill AI are doing exactly that. See the plans here →