Brand Voice Guide Builder: Tone of Voice Guidelines, Banned Words and AI Writing Rules, AI Skill for Claude & ChatGPT, Marketing Copywriting

Brand Voice Codifier is an AI skill for Claude and ChatGPT, made for marketers. Paste your best writing (and some that felt wrong). It extracts what makes your voice yours and builds an enforceable guide — sliders, banned words, paired examples, and the version-stamped enforcement block you paste into any AI. Nothing published yet? It builds from the customer emails and voice memos you already have.

Instant download. Works on the free tier. No coding. Your voice, turned into a paragraph any AI must obey.

WHAT IS AN AI SKILL?
Think of it as a trained assistant you add to Claude or ChatGPT. It is a small folder of files: the step-by-step method the AI must follow, the templates it fills in, the checks it runs on itself, and a worked example. You load it once (about a minute), then talk to the AI in plain English the way you would brief a colleague. Instead of a generic chat answer, you get the finished piece of work as a real file you can print, send or open in Excel. No coding, no software to install, nothing to configure. It runs inside your own Claude or ChatGPT account.

THE PROBLEM IT SOLVES
A draft comes back from a freelancer, or an AI, or a new hire. It's competent, it's on-message, and it's wrong, and the only feedback anyone can give is “make it sound more like us,” which helps nobody.

Voice feels unteachable because it lives in specifics: verb energy, sentence rhythm, how you hedge, the three words you would never use. Those things are describable. They're just rarely written down.

ABOUT THE EXAMPLES BELOW
The figures and names come from the built-in demo: Priya Shah, marketer at a Toronto bookkeeping software company, with a real GA4 export. Type "demo" after setup and you watch the same run happen in front of you before you type a single real number.

WHAT IT DOES (6 JOBS, ONE SKILL)
1. Codify the voice
Paste three to ten samples of your best writing, and a couple that felt wrong, because those teach more. Out comes a guide with sliders, banned words and paired examples.

2. Start with nothing published
No marketing copy you like? It harvests what you already have: the last emails you sent a customer, a five-minute voice memo, your messages to a colleague. You get a v0.9 hypothesis guide and a dated 30-day recheck that promotes it.

3. Write the bad-news voice
Outages, complaints, refunds, price rises. A separate block with the humor switched off rather than turned down, the hedging rules untouched, and a four-part structure to follow when nobody has time to think.

4. Cover the rest of the company
Support macros, sales follow-ups, in-product microcopy, job posts. Each gets its own register and its own rule: support acknowledges before it explains, error messages stay plain, sales still attaches proof to every claim.

5. Split founder and company
If a founder publishes under their own name too, you get two dialects with two complete blocks: same bans, same values, distinctly not the same person.

6. Audit for drift
Bring the guide and your last ten posts for a scored report, the three worst lines rewritten, and one row added to the ledger. By the third audit the ledger says whether the block is working or the rule itself is the problem.

WHAT YOU GET
• The enforcement block — one paragraph, any AI, your voice
• Ten “sounds like us / doesn't” pairs
• Variant blocks: founder, bad-news, second language

The guide is useful. The enforcement block is the product:
• The enforcement block, version-stamped, for any AI and any tool
• Variant blocks: founder, bad-news, second language
• Tone sliders, demonstrated with real rewritten sentences
• Ten “sounds like us / doesn't” pairs
• The 60-second test harness, and a one-page card for freelancers

YOUR FIRST FIVE MINUTES (copy-paste any of these)
› Here are 5 samples of our writing — codify the voice.
› We've barely published anything. Build from my customer emails.
› Build the sensitive-moment block for outage notices.
› Audit last month's posts for voice drift.
› demo — watch a bland corporate draft get its voice back.

HOW TO SET IT UP (about a minute)
1. Download the zip and unzip it. Inside: a START-HERE file, INSTALL steps for Claude, and CHATGPT-SETUP for ChatGPT.
2. Claude app: Settings → Capabilities → Skills → Upload skill (or add the files to a Project). Claude Code: drop the folder into ~/.claude/skills/. ChatGPT: follow CHATGPT-SETUP (copy-paste, about two minutes).
3. Type "demo" to watch the skill run on a built-in example. Then give it your own details in plain English.

WHO MADE THIS
The Skill Mastery builds AI skills for working professionals: photographers, filmmakers, creators, marketers and sales teams. We don't start from what AI can do. We start from how the job is actually done, the methods the trade already trusts (here: extraction-first repurposing, T-minus launch calendars and cost-per-result reporting), the numbers that have to be right, and the mistakes that cost real money. Then we engineer the skill so the AI follows those rules every single time, and we test it the way a buyer would use it: on the built-in demo, on a cold scenario it has never seen, and in a currency other than our own. Every skill ships with a demo mode, a self-check, and a plain list of what it can't do.

WHY NOT JUST ASK THE CHATBOT?
• It's portable, and it's stamped. The block works in ChatGPT, Claude, or whatever your team uses next year. Each one opens with the brand, the guide version and the date, so when three of them are circulating in someone's notes app you can tell which is current.

• It's testable. Most voice guides are documents nobody can verify. This one ships with a pass/fail harness: five voice points, plus a hard gate that fails the block outright if the output contains one invented fact.

• It bans invention. Every block carries a clause requiring each number, name, quote and date to come from the material in the cue, and to write [FACT NEEDED] instead of estimating. Style rules that quietly license fabrication are worse than no rules.

WHAT IT WON'T DO (read this before buying)
• A voice guide can't rescue a positioning problem. If what you're saying is unclear, saying it in your own voice just makes it clearly unclear.
• A guide built from emails and voice memos is a hypothesis, and it says so on every page until 30 days of real posts confirm it. It also works from the samples you give it: feed it your weakest writing and it will faithfully codify that.
• It makes AI write like you; a human still reads the words before they ship. A second-language block built without samples in that language is directional, and no automated harness replaces one native speaker reading it.

FAQ
Q: How is this different from just asking Claude or ChatGPT?
A: You can ask a chatbot to match a tone in one conversation. This produces a portable artefact: a version-stamped paragraph you paste into any AI, any time, plus a test harness that proves it works before you bet content on it, and an audit mode that catches drift months later.

Q: How many samples do I need?
A: Three to ten pieces of your best writing, and including two or three that felt wrong is genuinely useful, because the contrast teaches the boundary faster than good examples alone. If you have fewer than three, it builds from raw material you already have instead: customer emails, a transcribed voice memo, your messages explaining the product to a colleague. That guide ships labelled as a hypothesis with a dated recheck.

Q: What happens when something goes wrong and the jokes have to stop?
A: That's the sensitive-moment block: outages, complaints, refunds, price rises. Humor is switched off rather than reduced, the hedging rules stay exactly as they are, because pressure is when writers start softening, and the block follows a set structure. What happened, what it means for you, what we're doing, when you'll hear next.

Q: Our voice has changed since last year. Is the guide stale?
A: Audit mode answers that. It separates content drifting from the guide, which is fixed by re-issuing the block, from the voice genuinely evolving, which is fixed by updating the guide, because those need opposite responses. It also checks which version of the block the writers were actually using before it blames anyone's writing.

Q: Is it a subscription?
A: No. A one-time purchase, yours to keep, updates included.

Q: Do I need a paid Claude or ChatGPT plan?
A: No. The free tiers run it. Paid plans give you longer sessions and stronger models, which is nicer but not required.

Q: Claude or ChatGPT: which should I use?
A: Both are included and both work. Claude gives the fuller experience because skills upload directly; ChatGPT setup is a two-minute copy-paste from the included CHATGPT-SETUP file.

Q: Do I need to know how to code?
A: No. You type in plain English. Setup is upload-a-file, and every step is written out for non-technical people.

Q: Does it work outside the US, in my currency?
A: Yes. It works in your currency, your date format and your market. It never converts or invents exchange rates.

Q: Can I use it for client work?
A: Yes. One purchase is a perpetual single-user licence; everything it produces is yours to use, unrestricted, for your own business or your clients. You just can't resell or share the skill files themselves.

Q: Is my data private?
A: It runs entirely inside your own Claude or ChatGPT account. We never see your numbers, clients or files.

Q: Can I use it on my phone?
A: Set it up once on a computer; after that it works from the Claude and ChatGPT phone apps too.

Q: Do I get updates?
A: Yes, free updates to this version. If a new version ships, message us on Etsy for the update.

Q: What if it doesn't work for me?
A: Digital files can't be returned on Etsy, but message us and we'll make it right, whether that's a fix, a walkthrough or a refund. A real person replies.

PART OF THE GROWTH SUITE
Also in this shop: Content Repurposing Engine, Campaign Launch Kit, Report-to-Insight, or the full Growth Suite bundle at a saving. They share one profile, so you describe your business once and every skill reads it.

INSTANT DIGITAL DOWNLOAD
One .zip file, delivered by Etsy the moment you buy. Nothing physical will be shipped. Free updates for the current version. Questions before or after buying: message us here on Etsy, a real person replies.
Version 2.0.0.

AI disclosure: this toolkit was designed and written by The Skill Mastery with the help of AI tools, then tested by hand (every calculation, template and demo checked). It runs inside Claude or ChatGPT; it is not a standalone app. Not affiliated with Anthropic or OpenAI. Claude and ChatGPT are trademarks of their respective owners.