- A style guide exists to end recurring arguments, so build it from the disputes your team actually has rather than from a template of everything a guide could contain.
- Voice rules must be testable against a sentence. If a writer cannot check their draft against the rule, the rule is decoration.
- Adoption is designed, never hoped for: keep the guide to one searchable page, wire it into the review step, and assign one owner who updates it monthly.
Every content team above two writers eventually has the argument: serial commas, capitalization after colons, whether the brand says clients or customers, whether numbers under ten get spelled out. The argument repeats monthly, burns an hour each time, and settles nothing, because the settlement lives in the loser’s memory. An editorial style guide is where those settlements go to become permanent. This tutorial walks through building one that writers keep open in a pinned tab, which is a different artifact entirely from the sixty-page document most teams produce, admire, and never open again.
What a style guide actually fixes
The guide’s purpose is narrower than new teams assume, and the narrowness is the value. It does not make writing good; editing does that. It makes writing consistent, and consistency does three jobs. It removes the visible seams when four writers feed one blog, so the publication reads as one voice with one standard. It converts recurring debates into lookups, reclaiming the meeting hours that style arguments quietly consume. And it makes freelancers and new hires productive in their first week, because the answers to a hundred small questions exist in writing instead of in the heads of whoever has been around longest.
Consistency also reads as credibility from outside. Readers rarely notice a house style, but they notice its absence the way they notice a wobbling table: nothing they can name, everything slightly off. For a publication asking to be cited, that subliminal steadiness is part of the pitch.
Know what the guide cannot do, and refuse to let it try. It cannot fix unclear thinking, rescue a piece with nothing to say, or substitute for editing; a guide that attempts those jobs becomes a writing course bolted to a rulebook, and both halves suffer. Scope discipline at this stage decides everything downstream: the narrower the mandate, the more completely the document can fulfill it, and the more trust it accumulates every time someone opens it and finds an answer instead of a lecture.
What belongs in yours
The strongest guides are short, opinionated, and local. National newspapers need exhaustive manuals; a marketing blog needs seven sections and the discipline to stop there.
Two sections do the most work. The word list is the one writers open daily: house spellings, product names with correct casing, the terms your industry disputes, and the words you have banned with their approved replacements. The escalation section is the one that keeps the guide short, and the professionals model it well. Google’s public developer documentation style guide resolves anything it does not cover by deferring outward, first to project style, then to the guide itself, then to named third-party references for spelling and general usage. Your version can be one paragraph: what the guide is silent on, a dictionary decides; what the dictionary is silent on, the editor decides; and every editor decision that repeats gets written in.
The word list
Seed the word list by mining your last twenty published pieces for inconsistencies: the product referred to three ways, the varying capitalization, the competitor spelled differently across posts. That audit takes an hour, surfaces the disputes that are real rather than hypothetical, and hands you the first thirty entries. The list then grows one line at a time, every time an editor corrects the same thing twice.
Give banned words their own visible block within the list, each with its approved replacement, because prohibition without alternative just moves the problem to the writer’s deadline. If leverage as a verb is out, say what is in. If world-class is banned, show how the guide wants superiority claimed instead, or whether it wants it claimed at all. The replacements teach the voice faster than the voice section does, one small concrete verdict at a time.
Defining voice writers can use
Voice sections fail in a predictable way: they describe a personality instead of constraining a sentence. Friendly but professional, confident yet humble, expert but accessible; every brand on earth claims the same four adjectives, and no writer has ever checked a draft against them.
The fix is to write voice as tests a sentence can pass or fail. Name a reference person: a senior colleague explaining to a smart newcomer, a practitioner writing to practitioners. List the moves your voice makes and refuses: contractions welcome, hype adjectives banned, claims carry evidence or hedges, jokes allowed in introductions and never in instructions. Then anchor it with pairs: three sentences in the house voice beside their off-voice twins, with one line on the difference. Writers imitate examples far more reliably than they interpret adjectives, and the pairs section will be the most-read part of the guide.
Add one paragraph on register shifts, because voice is not one setting. The blog post, the error message, the sales one-pager, and the apology email share a speaker and should not share a mood. Two lines per content type, mostly about what gets dialed down, keeps a single voice from flattening into a single register, which is the subtle way consistent brands turn monotonous.
Choosing rules worth writing down
Guides die of ambition. The author, enjoying the work, documents every preference they have ever held, and the result is a manual too long to search and too fussy to respect. Admission needs a standard, and two questions supply it.
The disputed-twice test keeps out rules that solve imaginary problems: if nobody has argued about it, editors are handling it, and the guide adds weight without adding order. The reader-notice test keeps out private fussiness: rules about matters invisible to readers spend the team’s patience on nothing. What survives both filters is a short list of consequential, contested calls, each written as a flat ruling with one example, because a guide that explains its reasoning at length becomes an essay, and writers under deadline pressure do not consult essays. Even Google’s guide opens by granting permission to break its rules when clarity wins, quoting Orwell to make the point; a guide is a floor for consistency, never a ceiling on judgment.
One edge case needs handling explicitly: house rulings that contradict common usage, such as a product name spelled against convention. Those get a one-line reason in the guide, because a rule that looks like a typo gets silently corrected by the next well-meaning editor.
Write the rulings in the imperative and resist the urge to soften them. Use serial commas. Spell out numbers one through nine. Capitalize the product, lowercase the category. Hedged rules, where writers are invited to consider perhaps preferring an approach, reproduce the ambiguity the guide exists to end. The kindness is in the brevity: a writer on deadline wants a verdict in five seconds, and the guide that delivers verdicts gets opened again.
Writing it in a week
The sixty-page guide takes a quarter and dies on arrival. The usable version takes five working sessions:
- Day one: run the twenty-piece audit and collect every inconsistency and past dispute into a raw list.
- Day two: rule on the mechanics, spelling standard, numbers, dates, capitalization, and punctuation, each in one line.
- Day three: draft the voice section with its reference person and three example pairs.
- Day four: build the word list and the formatting rules for headings, links, lists, and visuals.
- Day five: write audience, sourcing, and escalation, then have the two most senior writers attack the draft.
Ship it at perhaps eight pages, dated, with a changelog line at the top. The brevity is a feature to defend in review: every addition proposed after launch faces the same two admission questions as the original rules, which is how the guide stays the length people read.
Expect the first month to generate amendments, and treat that as the system working. The audit cannot catch every live dispute, so the early weeks surface the misses: a term two writers still spell differently, a formatting case the guide never imagined. Each one gets a ruling, a line in the changelog, and the guide hardens from plausible to authoritative in about six weeks of ordinary publishing.
Getting it used
Publication is the halfway point for guides too. The document nobody opens enforces nothing, and adoption is a design problem with known solutions. Keep it one searchable page rather than a PDF, because writers will search it or nothing. Link it from the content brief so every assignment starts one click from the standard. Give it an owner who logs rulings monthly, because an unowned guide freezes at its launch date and thaws into irrelevance within a year. Announce updates in one line wherever the team coordinates, because a silent changelog is a private diary.
Enforcement without resentment
The guide earns authority at the review step or nowhere. Editors should cite it by section when correcting, which converts corrections from personal taste into shared law and spares everyone the relitigation. Freelancers get the guide with their first brief and a note that review checks against it; guest contributions get held to the same standard, since editors on the receiving end apply the same tests to guest posts they accept. Teams without a standing editor can borrow one: our editorial review service enforces house style across everything that ships, guide in hand. However the enforcement is staffed, the loop is what matters: the guide informs the review, the review feeds the guide, and the arguments that used to repeat monthly become a changelog line nobody has to sit through.
A practical tell that the guide is working is the shape of review comments. Before it, editors leave taste notes the writer can argue with. After it, they leave section numbers, and the arguments stop.
Measure the guide the way you would measure any tool: by the arguments it ends. If the same dispute survives three months of the guide existing, the relevant ruling is missing, unfindable, or unenforced, and each of those has a different fix. A guide that is working shows up as an absence, fewer style comments in review, faster freelancer ramp-ups, and a blog that reads like one careful person wrote all of it. That absence is the deliverable, and it compounds with every piece shipped under it.
Want an editor who enforces the standard so the guide does not have to?
