Semicolon frequency analyser

By Jack Stovell · published 2026-09-24

Semicolon frequency (semicolon_frequency) measures how often a writer uses semicolons, as semicolons per 1,000 words. It is one of 45 attributes in a ScriptGrain voice profile, in the Syntax group. Across 299 published pieces the median is 0.45 per 1,000 words, and at least a quarter of pieces use none.

Analyse your semicolon frequency free

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

Try the six-signal check first

At a glance

What semicolon frequency measures

Semicolon frequency measures exactly what it says: the rate at which semicolons show up in your writing, per 1,000 words rather than as a raw total. That per-1,000-words scale matters because it makes a 400-word blog post and a 4,000-word report comparable. The app's own definition keeps it plain: "how often semicolons are used." In the API it's the key `semicolon_frequency`; in the app it shows up as "Semicolon Frequency" under the Syntax group, alongside things like average sentence length, sentence length variance and clause ordering.

A voice profile is built by an extraction model (Claude Sonnet 5) that reads your samples in two passes, one per sample, then a synthesis pass that reconciles everything into a single profile. So the number stored against your name is that model's reading of your writing, not a live count updated every time you type. When code scores a draft, or when you run the free writing style analysis or the habits report, it counts semicolons per 1,000 words directly from the text; that's a separate count sitting alongside the profile's own extracted value. Two related counts use other units: the AI cliché checker flags "semicolon chains" (three or more semicolons in one sentence), and the homepage's "Punctuation" signal counts semicolons and colons together per 100 words.

How semicolon frequency is scored

Semicolon frequency is scored as part of voice match, at a weight of 1.5. Code measures your draft's actual rate (via `measureTextFeatures`), compares it to your profile's stored rate, and works out similarity as one minus the gap divided by a tolerance, floored at zero so a wildly off draft can't score negative. Tolerance is the larger of 2 per 1,000 words or your own profile's rate, so a habitual semicolon user gets more room to vary than someone who almost never uses one. There's a rare-habit rule too. If both your draft and your profile sit under 0.5 per 1,000 words, agreement only counts at half weight and is never treated as a headline match, because plain text with zero semicolons technically "agrees" with a profile that barely uses them, and that's not the same as a genuine stylistic fingerprint. For the full mechanics of how this and 27 other attributes feed into the overall score, see the voice measurement framework and the voice match glossary entry.

How to analyse semicolon frequency in your own writing

Analysing semicolon frequency by hand takes about five minutes with a sample you trust.

  1. Pull a writing sample of at least a few hundred words; longer samples smooth out flukes.
  2. Count every semicolon in the piece. Don't count colons or em dashes; only the semicolon character itself.
  3. Count the total words in the sample.
  4. Divide the semicolon count by the total word count, then multiply by 1,000. That's your rate per 1,000 words.
  5. Repeat across two or three samples and average the results, since one unusually semicolon-heavy paragraph can skew a single piece.
  6. Check for chains: any sentence with three or more semicolons is worth flagging on its own, regardless of your overall rate.

ScriptGrain's free voice profile reads semicolon frequency from a writing sample among all 45 attributes.

Typical semicolon frequency in published writing

Typical semicolon frequency in published writing is low: the median is 0.45 per 1,000 words. ScriptGrain's reference corpus is 299 recent public pieces from 47 sources (blogs, newsletters and essay sites), built on 20 September 2026 and measured with the product's own code. Its tenth percentile sits at 0 semicolons per 1,000 words, the 25th percentile is also 0, the median is 0.45, the 75th percentile is 1.59, and the 90th percentile is 3.13. So at least a quarter of the pieces use no semicolons at all, half use 0.45 or fewer per 1,000 words, and only the top tenth reach 3.13. A steady semicolon habit is unusual, which makes it distinctive.

How to make AI use semicolons like you

Getting an AI assistant to use semicolons the way you do takes deliberate instruction, not a hope that it'll pick up the habit on its own.

  1. Tell it directly: allow a semicolon between two closely linked independent clauses, and nowhere else.
  2. Show it examples: two or three of your own sentences that use semicolons the way you do.
  3. Cap it at one semicolon per sentence.

When a ScriptGrain profile generates text for you, semicolon frequency travels inside the voice profile JSON sent to the model, under an instruction to treat every attribute as a hard constraint rather than a suggestion. The generation step also bans one thing outright, regardless of your own habits: four or more independent clauses chained together with semicolons in a single sentence. The profile reaches ChatGPT, Claude and other tools through the API and the MCP server.

How to make AI use fewer semicolons

Making an AI assistant cut back on semicolons is mostly a matter of giving it somewhere else to put the pause.

  1. Replace semicolons with full stops. A semicolon joins two clauses that could each stand alone, so a full stop always works.
  2. Break any chain of semicolon clauses into separate sentences entirely; don't just remove one semicolon and leave the rest.
  3. Use a colon only where the second clause explains or unpacks the first.

A ScriptGrain profile with a low semicolon frequency sends that rate as a hard constraint on every draft, and its generation prompt always bans four or more clauses chained with semicolons. The profile reaches ChatGPT, Claude and other tools through the API and MCP server; for the rest of your punctuation habits, see the punctuation and format reference and parenthetical rate.

Questions

How often should you use semicolons?

There's no universal correct rate, but the reference corpus gives a useful benchmark: median use across 299 published pieces is 0.45 semicolons per 1,000 words, and at least a quarter of the pieces use none. At 3.13 or more per 1,000 words you're in the top tenth, which is fine, but worth being deliberate about rather than accidental.

How is semicolon use scored?

It's scored as part of voice match with a weight of 1.5. Code counts your draft's actual rate, compares it to your profile's stored rate, and calculates similarity as one minus the gap over a tolerance (the larger of 2 per 1,000 words or your profile's own rate), floored at zero. A rare-habit rule halves the weight of the agreement when both sides sit under 0.5 per 1,000 words, and never headlines it as a match.

How do you make AI use fewer semicolons?

Tell it plainly: replace semicolons with full stops, break up any sentence chaining multiple clauses with semicolons, and reserve colons for cases where the second clause explains the first. Give it that as an explicit instruction rather than a vague "write naturally" prompt, then count semicolons per 1,000 words in the output against your own rate.

How do you make ChatGPT stop using semicolons?

Give it a hard cap, not a preference. Tell it to use full stops instead of semicolons, to never chain three or more semicolon clauses in one sentence, and to use colons only when the second half explains the first. ScriptGrain's own drafts always ban four or more clauses chained with semicolons and aim at your profile's rate; ChatGPT can read that profile through the API and MCP server.

Are semicolons a sign of AI writing?

Not on their own. Plenty of human writers use semicolons, and the median piece in ScriptGrain's reference corpus uses 0.45 per 1,000 words. ScriptGrain's AI cliché checker flags one pattern, the semicolon chain (three or more semicolons in one sentence), and its generation prompt bans four or more clauses chained that way. A count describes a habit; it cannot say who wrote the text.

Related

Sources