A Living Style Guide for Consistent AI Rewrites
Without a memory, every AI rewrite starts from zero. It doesn’t know your brand voice, your preferred terms, the things you always do and never do — so the results drift: a little formal here, a little generic there, your product name spelled three ways across a week. A living style guide fixes that by giving the editor persistent rules and voice that get applied to every edit, automatically. This page explains how that memory works, why it matters more than any single prompt, and how to build one.
Why one-off rewrites drift
Each isolated rewrite is a fresh request with no history. The model does a reasonable job in that moment, but “reasonable” varies, and across many edits the variation compounds:
- Voice wobbles. One rewrite comes back crisp and confident; the next, hedged and corporate — because nothing told the model which one you are.
- Terminology splinters. Your feature is “Smart Sync” but rewrites variously produce “the sync feature,” “auto-sync,” and “Smart-Sync.” Small, but it reads as careless across a body of work.
- Banned patterns sneak in. The em-dash habit you hate, the exclamation points your brand avoids, the jargon you’ve sworn off — without a rule, they reappear.
- Audience register slips. You write for non-technical buyers; an un-guided rewrite reaches for engineering vocabulary.
You can fix each rewrite by hand or by stuffing a long instruction into every prompt — but that’s exactly the friction inline editing was supposed to remove. The scalable fix is to say it once and have it applied every time.
What a living style guide is
A living style guide is a stored set of voice rules and facts that the editor injects into every rewrite as context, so the model edits with your standards already in hand. It typically holds:
- Voice and tone — “Confident but warm. Plain language. No corporate filler. Short sentences over long ones.”
- Terminology — the exact spelling and casing of product names, features, and key terms, plus preferred phrasings (“say teammates, not users”).
- Do/don’t rules — “No exclamation points. Avoid ‘leverage,’ ‘utilize,’ ‘seamless.’ Oxford comma always.”
- Audience — who you’re writing for, so register stays consistent.
- Examples — a few snippets of on-voice writing the model can pattern-match to.
“Living” because it’s not a one-time setup — you refine it as you notice the AI getting something wrong, and the correction sticks for all future edits. Over time it converges on your voice.
Style guide vs. context-awareness
These two work together but solve different problems:
- Context-aware rewriting reads the text surrounding your selection so the edit fits this document — the local tone, the terms already on the page.
- A living style guide carries your voice and rules across every document, regardless of what’s around the selection.
Context makes a rewrite fit the paragraph; the style guide makes every rewrite sound like you. The best output uses both: your durable voice, adapted to the local document.
How a style guide relates to presets
A living style guide and custom prompt presets layer cleanly. Presets define the job (“make professional,” “summarize”); the style guide defines the voice every job is performed in. So your “make professional” preset doesn’t just produce a professional version — it produces your professional version, with your terms and rules, every time. You write the voice once; it flavors every preset and every ad-hoc edit thereafter.
Building one that works
- Start small and specific. A handful of concrete rules (“no exclamation points,” “say teammates not users,” “plain language”) beats a vague paragraph about “our brand essence.”
- Add the terms that get mangled. Product and feature names are the highest-value entries — they’re the errors readers notice.
- Refine from real misses. Each time a rewrite gets your voice wrong, add the rule that would have prevented it. That’s the “living” part.
- Give it an example or two. A short on-voice sample teaches the model faster than a list of adjectives.
A living style guide in EditSnappy
EditSnappy can carry a persistent style guide — your voice, terms, and rules — and apply it to every rewrite, so edits sound like you instead of generic AI, without re-typing your standards into a prompt each time. It works hand in hand with EditSnappy’s context-awareness (which fits the edit to the current document) and your custom presets (which define each job), all running through the same safe inline loop: the change streams into place, shows as a diff before it commits (Tab to accept, Esc to keep your original), preserves your formatting, and is undoable with one key. Same behavior on Mac and Windows.
[[MISSING: confirm the exact shape of the style-guide / persona-memory feature with Ken — master-sales-copy §5 lists it as a reach goal from the white space, so verify scope before stating specifics in marketing.]]
Say your voice once. Get it on every edit. See how it works on the homepage →