House Style Guide Template

Don't rewrite rules a published guide already settles. Make one your default, usually AP for news and business writing or Chicago for books. Google's developer documentation style guide sets out the order to check¹: a project's own style first, then Google's guide, then Merriam-Webster for spelling and Chicago for nontechnical style.

Name the edition too, because base guides change. The 58th edition of the AP Stylebook has more than 200 new and revised entries², including healthcare and primetime as single words. Chicago's 18th edition, from 2024, now endorses singular they³ as an option for a person whose gender is unknown.

Russell Harper, principal reviser of two editions of The Chicago Manual of Style, wrote in 2020 that consistency is second only to being right⁴ among editorial principles. So when neither the base guide nor the dictionary settles a question, decide once and write it down. Most entries fall into five groups.

• Every departure from the base guide, flagged as a departure
• Your own names: the organization, products, programs, places and people's titles
• Formats the base guide skips, such as headlines, captions, chart labels and buttons
• Terms your readers care about, with the reason for your choice
• Cross-references from the words people search to the entry that answers them

Test any existing style sheet before you change it. Pull a sample of recent work across writers and formats. For each rule a text search can check, count every place it applies and how often the copy followed it. GOV.UK, the UK government's website, uses the same evidence when style guides run out: it tells publishers to compare how often different spellings or formats are used⁵ on GOV.UK.

The common mistake is counting too much. Measure a rule on numerals for ages against every age in the sample, not every number. Then rate each rule on its own, because an overall figure hides the failures. In the template's made-up newspaper, the copy follows the sheet 86.7% of the time, yet 14 rules are followed under 60% of the time.

For each weak rule, count how many different writers break it. If most of the team does, your writers have voted and the rule is probably wrong. If two or three people do, that is a conversation with them, not a rule change.

Questions show what the guide is missing. Log every style question in one place, from chat, email or comments on drafts, with the date and who asked. These records are easy to lose, and they are your best evidence.

1. 1

   Group them

   Merge different wordings of the same question. In the template's made-up newsroom, 487 queries came down to 96 questions.

2. 2

   Rank them

   Sort by how often each was asked. There, the top 12 questions made up 47% of all queries.

3. 3

   Match each to an entry

   Note whether it has no entry, an entry your published copy contradicts, or an entry filed under a word nobody searches.

4. 4

   Rule from the top

   Write entries for the most-asked questions first, and stop when the questions thin out.

Expect some answers to depend on the reader. AP's editors field subscribers' style questions and publish a sampling of the answers⁶. When a writer whose agency builds on AP asked about IT, the editors said general readers need information technology spelled out the first time, but technical readers don't.

File each entry under the word people search for, with cross-references from the other words they try ("DA, see district attorney"). Give the rule in one sentence, a right and a wrong example, and where it applies: body copy, headlines, captions or charts. Say whether it departs from your base guide, and date every ruling.

Then give the reason, written for the writer who has done it the other way for years. A reason about your readers ends arguments that "AP says so" only starts. GOV.UK's A to Z prefers "page" to "screen", for example, because user research by the Government Digital Service⁷ noticed that users understand "page", even in a mobile app.

Answer each question once, in a register of rulings, and point people to the row instead of replying privately. Name one owner who rules. When someone disagrees, ask for their reason, then reopen the row and keep both decisions on it. GOV.UK asks the government publishers who suggest changes to its guide to say why and to include evidence, for example user research or data⁵.

Recheck your rulings whenever the base guide publishes a new edition, and rerun the sample once a year. Give new writers one page: the 10 to 15 most-asked questions and every departure from the base guide. They will absorb the settled rules by reading what you publish.

A style guide decides how to write a name, not whether a claim is true, so run fact-check comments on your draft as a separate pass. The template on this page has the guide, the register and the measuring sheets. Open it in River, which tests your style sheet against your published work for you.