Capitalisation checker

By Jack Stovell · published 2026-09-24

Capitalisation quirks is the attribute in a ScriptGrain voice profile that records a writer's capitalisation habits, such as sentence case headings or capitals for emphasis. The extraction model writes it as a short description of up to 15 words. It is judged, not counted, and it never moves the voice match score.

Analyse your capitalisation quirks free

Your free voice profile measures capitalisation quirks among all 45 attributes, from a sample of your own writing. No card needed.

Try the six-signal check first

At a glance

What capitalisation quirks measures

Capitalisation quirks measures the habits a writer shows around capital letters: whether headings run in sentence case or Title Case, which brand names or terms get capitals every time, and whether capitals ever get used for emphasis instead of italics. The app's own definition is blunt: "Any noted capitalisation habits." No formula, no count, just an observation written in plain English.

This is a judged attribute, not a counted one. It sits in the punctuation and format group, one of the 45 attributes, alongside comma density and question marks in body, which code counts when a draft is scored. It's a description, stored as a string of at most 15 words. Think of it as the model's note to itself; example values: "this writer title-cases every heading" or "capitalises Brand Name consistently, nothing else stands out."

That distinction matters for anyone comparing attributes across a profile. Some are numbers that can be set against a reference corpus. This one is a read, closer to an editor's margin note than a spreadsheet cell.

How capitalisation quirks is scored

Capitalisation quirks is not scored: the voice-match engine has no comparator for it, so it never affects your voice match score. Here's how it gets into your profile instead: Claude Sonnet 5 reads each writing sample you upload in Pass 1, noting capitalisation behaviour where it appears. Pass 2 synthesises those notes across all your samples into one description for the profile.

So it's judged, the same way formality or humour register gets judged. With no comparator, it sits outside the 28 attributes that can move your score, and polish can't target it either: there's no reference value to close the gap against.

That doesn't make it decorative. It still goes to the model with every ScriptGrain draft; it just doesn't factor into the percentage you see on a scorecard. The voice measurement framework covers which attributes do move the needle, and the voice match glossary entry explains the score itself.

How to analyse capitalisation quirks in your own writing

You can spot your own capitalisation habits by hand in a few minutes, no tool required.

  1. Pull up three or four pieces you've written recently: emails, blog posts, whatever's typical.
  2. Look at every heading and subheading. Are they in sentence case ("How to write better headings") or Title Case ("How To Write Better Headings")? Pick whichever shows up most.
  3. Scan the body text for words you capitalise that don't strictly need it: role titles, internal jargon, product names.
  4. Note anything you capitalise purely for emphasis, like shouting a single word in the middle of a sentence.
  5. Check brand and product names. Are you capitalising them the way the brand actually writes them, or your own version?

Write down what you find in 15 words or fewer, the same limit the profile uses. ScriptGrain's free voice profile reads this habit from a writing sample among all 45 attributes.

Examples of capitalisation quirks in real writing

Capitalisation quirks show up in small, specific ways once you know where to look.

Example: a newsletter that titles every section header in full Title Case, like "Why Your Emails Are Getting Ignored", while the body copy stays in ordinary sentence case throughout.

Example: a brand voice guide that always capitalises the product name in full, "SmartLedger", but never capitalises generic terms like "software" or "dashboard" elsewhere in the same document.

Example: a blog post that capitalises a single word mid-sentence for emphasis, such as "this is the ONE thing that matters", rather than using italics or bold.

Three different habits, three different writers. None of them wrong, just distinctive.

How to make AI keep your capitalisation habits

Keeping your capitalisation habits in AI-generated text starts with telling the model exactly what those habits are, because it can't guess them.

  1. State plainly whether you use sentence case or Title Case for headings. Don't leave it implied.
  2. List the specific words you always capitalise and the ones you deliberately never do.
  3. Give the model one real sample where the habit is clearly visible, ideally a heading and a paragraph together.

When a ScriptGrain profile generates text, capitalisation quirks travels inside the voice profile JSON block of the system prompt, under "Treat every attribute as a hard constraint, not a suggestion." There's no separate rule written just for capitals; it rides along with everything else. The same profile reaches ChatGPT, Claude and other tools through the API and MCP server.

How to make AI use standard capitalisation

Making an AI assistant use standard capitalisation means asking for it directly and stripping out anything that contradicts it.

  1. Ask explicitly for sentence case headings and standard capitalisation throughout.
  2. Remove capitals used for emphasis, both in your instructions and in any sample you give it.
  3. Keep brand names capitalised exactly as the brand itself writes them, nothing more, nothing less.

A ScriptGrain profile sends whatever capitalisation note the extraction model wrote from your samples, as a hard constraint like every other attribute, and the same profile reaches ChatGPT, Claude and other tools through the API and MCP server. For a neighbouring punctuation habit, see exclamation rate.

Questions

What is a capitalisation checker?

A capitalisation checker, in ScriptGrain's sense, isn't a grammar tool that flags errors. It's the part of a voice profile that notes your capitalisation habits, sentence case versus Title Case, which words you always capitalise, whether you use capitals for emphasis. It's read from your writing samples by the extraction model, stored as a short description, and used when generating new text in your voice.

What's the difference between sentence case and Title Case?

Sentence case capitalises only the first word (and proper nouns) of a heading, like "How to write better headings". Title Case capitalises most major words, like "How To Write Better Headings". Neither is objectively correct; it's a style choice. What matters for voice consistency is picking one and applying it consistently, then telling any AI tool which one you use so it doesn't default to its own habit.

How is capitalisation measured in writing style?

It isn't measured with a formula or a count. An extraction model reads your samples and judges your capitalisation habits, writing a short free-text description. When ScriptGrain scores a draft or runs its free tools, code counts 14 of the 45 attributes, and this is not one of them. It is judged, it has no comparator, and it doesn't move a voice match score.

How do I make ChatGPT use sentence case?

Tell it directly: "use sentence case for all headings, capitalise only the first word and proper nouns." Give it one example heading written the way you want. Make the instruction explicit rather than assumed. A ScriptGrain profile carries the habit only if the extraction model noted it in the capitalisation quirks field, so the direct instruction is still worth giving.

How do I stop AI capitalising every word in headings?

State your preference plainly (sentence case), then show one example. Vague instructions like "keep it casual" won't reliably stop Title Case defaults; models need a direct rule. If you use a ScriptGrain voice profile and your samples show sentence case headings, the extraction model can note it in the capitalisation quirks field, which goes to the model as a hard constraint with every ScriptGrain draft.

Related

Sources