Brand Voice Guide is a Claude AI skill — a structured voice document Claude reads before writing anything for your brand.
Read fifty LinkedIn posts written with AI assistance and you can feel the sameness before you can name it. The rhythm. The measured enthusiasm. The opening hook that gestures at a relatable problem. The three-part structure. The closing line that pivots to an invitation. It's not bad writing — it's competent, legible, structurally sound. It's also completely interchangeable. Pull the name off the post and you couldn't identify the author. That's the problem.
Most people blame the model. Claude sounds like ChatGPT sounds like Gemini — they're all doing the same thing. But the homogeneity isn't coming from the model. It's coming from the absence of context. When Claude doesn't know who you are, what you sound like, or who you're writing for, it writes at the median. The median is technically correct. The median is also forgettable.
The fix people usually try is better in-session prompting. "Write this in a more conversational tone." "Make it punchier." "Less corporate." These instructions move the needle a few degrees — but they're corrective, not preventive. You're steering Claude away from the wrong thing rather than pointing it at the right one. And you're doing it every single session, from scratch, because Claude has no memory of the corrections you made last time.
Why "Write Like Me" Doesn't Work
The instinct to paste in an example and say "write more like this" is understandable. Claude is good at pattern-matching — give it a sample of your writing and it can approximate the surface features. The problem is that surface features aren't voice. They're the output of voice. What creates a consistent, recognisable brand voice is a set of decisions that happen before a word is written: who you're writing for, what they already know, what you never say, how you structure an argument, where you put the point.
Claude can't reverse-engineer those decisions from a single example. It can pick up your paragraph length and your vocabulary register. It can't pick up that you never use rhetorical questions because you find them condescending. It can't pick up that your audience is sceptical of optimism and responds to realism. It can't pick up that your natural structure is claim-evidence-implication, never problem-solution-CTA.
Telling Claude to "write like this example" gives it your surface features. A structured voice document gives it your decision rules — which is what actually produces consistent output across every task, every session.
This is why the editing never ends. Claude gets close enough to be usable but never close enough to be yours. You spend ten minutes per post pulling the output back into something you'd actually publish. Multiply that across a week of content tasks and you've lost the time AI was supposed to give you — correcting for missing context rather than spending that time on work that requires your judgment.
That gap is exactly what the Brand Voice Guide skill for Claude was built to close.
What a Real Voice Document Contains
A brand voice document that Claude can actually use isn't a list of adjectives. "Direct, warm, conversational" describes approximately forty percent of all brands — including ones with completely opposite tones. Adjectives are aspiration. Claude needs operation.
An operational voice document answers five specific questions, the same ones Claude asks silently every time it writes anything:
What are the decision rules for tone? Not "be direct" — "put the conclusion in the first sentence, every time, without exception." Not "be warm" — "acknowledge the reader's situation before offering a solution; never skip straight to the fix."
What are the banned patterns? The words, phrases, and structures your brand never uses — and what you use instead. The substitution is as important as the prohibition. Claude needs to know what to reach for when the banned option is the natural one.
Who is the reader? Not a demographic — a person. What do they already know? What are they sceptical of? What do they want to be told, and what do they find patronising? Claude adjusts its entire register based on assumed reader expertise and assumed reader cynicism.
How is writing structured? Paragraph length, heading use or absence, whether you use lists or prose, how arguments are sequenced. These structural patterns are often more responsible for voice than the vocabulary choices people focus on.
What does the transformation look like in practice? Before/after rewrites showing generic output converted to your voice — because examples do more work than rules. Claude learns the pattern more precisely from watching the conversion than from reading the description of it.
The difference between prompting for tone and having a voice document is the difference between giving someone directions and giving them a map.
The Sessions That Still Go Wrong — and Why
Even with a good voice document, three things will still produce generic output. It's worth knowing what they are so you can avoid them rather than blame the document.
The brief is too vague. "Write a LinkedIn post about our new feature" gives Claude almost nothing to work with beyond the voice guide. The more specific the brief — the specific audience for this post, the one thing it should make them feel, the specific claim it should land — the more the voice guide can do its job. Voice fills the how. The brief still has to fill the what.
The voice document is describing aspiration, not current reality. If you write your voice guide to describe how you want to sound rather than how you actually sound, Claude will write a version of you that doesn't exist yet. Build the guide from real content you've published, not the brand you intend to be.
The session is too long. Claude's attention to early context degrades over a very long conversation. If you establish voice at the top of a session and then spend an hour going back and forth on unrelated tasks, the voice calibration weakens. For content-heavy sessions, re-paste the relevant section of the guide at the start of each major task block.
What Changes When Context Is Right
The shift isn't dramatic — it's precise. Claude still does the structural work. It still generates the draft. What changes is that the draft lands in your register instead of the average register, which means the editing is refinement rather than rescue. You're adjusting emphasis and sharpening a claim rather than rebuilding tone from scratch.
The first is Claude writing to the median — technically competent, structurally correct, and completely without personality. The second is Claude writing with an audience profile (sceptical practitioners, not aspirational beginners), a structural rule (claim first, solution second), and a banned pattern (no rhetorical warmth, no exclamation points). Same topic. The document is what separates them.
The AI content problem most people are trying to solve with better prompts is actually a documentation problem. The voice exists — in your head, in your published work, in the editing you do every time Claude gets it slightly wrong. The Brand Voice Guide skill's job is to get that out of your head and into a format Claude can use. Once it's there, the prompts get shorter, the editing gets lighter, and the output starts sounding like someone actually wrote it — because, in every way that matters, someone did.
If you're working across the full Creator workflow, the Creator bundle covers everything in one place.
Put this to work: the Brand Voice Guide skill for Claude turns everything above into one guided workflow you run in a normal Claude chat. Not ready to buy? Start with a free Claude skill and see how it works first.