Article usage analyser

By Jack Stovell · published 2026-09-24

An article usage analyser measures how your uses of "the", "a" and "an" split between the three, as three shares. ScriptGrain stores it in a voice profile as the_a_an_ratio, one of five Function words attributes, and weights it 0.75 in the voice match score. Nobody sets this balance on purpose, so it works as a marker.

Analyse your article ratio (the, a, an) free

Your free voice profile measures article ratio (the, a, an) among all 45 attributes, from a sample of your own writing. No card needed.

Try the six-signal check first

At a glance

What article use (the, a, an) measures

Article use measures the balance between "the" (definite) and "a" or "an" (indefinite) across your writing. Articles are function words: small words that carry grammar more than meaning. The scale is three shares, one each for "the", "a" and "an"; scoring normalises them so they sum to one. Write mostly about specific, named things ("the report", "the client") and your "the" share climbs. Write mostly about general instances ("a customer", "an invoice") and your indefinite shares grow instead.

Nobody decides between "the" and "a" for effect, and the app's own definition calls the ratio "an unconscious stylistic marker". It's closer to a fingerprint than a technique, which is why it sits in the Function words group with the other small words you use without thinking.

The value stored in a voice profile comes from Claude Sonnet 5 reading your writing samples, not from you telling the app anything about your habits. Pass 1 reads each sample and returns every attribute, including this one; pass 2 synthesises a single profile across all your samples. So the shares you see are the extraction model's reading of your writing. Code counts the three articles only when a draft is scored against the profile.

How article use (the, a, an) is scored

Article use is scored, with a weight of 0.75 in the voice match score. When a draft is scored against a profile, code counts its "the", "a" and "an" directly from the text, normalises both sets to shares, then compares them. The similarity score is one minus half the summed absolute difference between them.

In practice, the closer a draft's article balance sits to your profile's, the higher this component scores. Get it wildly wrong (all "the", no "a") and the gap drags the number down. Twenty-eight of the 45 voice attributes can move the overall score, and this is one of them.

For the full mechanics of how individual attributes combine into one voice match figure, the voice measurement framework walks through it end to end.

How to analyse article use (the, a, an) in your own writing

You can do this by hand. It's tedious but not difficult.

  1. Take a writing sample of at least a few hundred words, ideally something typical of your usual work.
  2. Count every occurrence of "the".
  3. Count every occurrence of "a" (as a standalone word, not part of another word).
  4. Count every occurrence of "an".
  5. Add the three counts together to get your total article count.
  6. Divide each individual count by the total to get three shares.
  7. Repeat across two or three samples and compare; consistency across samples tells you it's a genuine habit, not a one-off.

ScriptGrain's free voice profile reads it from a writing sample, alongside the other 44 attributes.

Examples of article use (the, a, an) in real writing

Example: "The report landed on the director's desk before the meeting." Every noun here is definite; the writer assumes the reader already knows which report, which director, which meeting.

Example: "A customer called about an invoice she'd never received." Both nouns are indefinite; these are instances, not named things the reader's supposed to recognise.

Example: "The team reviewed a proposal from a new supplier." This one mixes both: "the team" is specific and known, "a proposal" and "a new supplier" are being introduced for the first time. Most real writing sits somewhere between the first two examples, leaning one way or the other depending on subject and habit.

How to make AI refer to specific things like you do

If your writing leans towards "the" over "a" and "an", here's how to get an AI assistant to match that.

  1. Refer to things the reader already knows by name: the report, the team, the deadline.
  2. Favour specific instances over general categories when you're choosing what to write about.
  3. Give the AI real samples of your own writing so it can copy the pattern rather than guess at it.

A ScriptGrain profile sends your article shares in the voice profile JSON of every generation prompt, under the instruction "Treat every attribute as a hard constraint, not a suggestion." There is no separate rule for articles. The same profile reaches ChatGPT, Claude and other tools through the API and MCP server, so the habit travels with you.

How to make AI write about things in general like you do

If your writing leans towards "a" and "an", the steps run the other way.

  1. Introduce new examples as instances: a customer, an invoice, a supplier.
  2. Write about kinds of things as well as specific named ones.
  3. Give the AI real samples so the general-instance habit gets copied, not invented.

The same profile block works in both directions: a ScriptGrain profile carries your article shares into every draft it writes, and into ChatGPT, Claude and other tools through the API and MCP server, so your samples decide which way the drafts lean. Its sibling pronoun distribution is stored and scored the same way, as shares of I, we and you.

Questions

What is an article usage analyser?

It's a way of measuring how often you write "the" against "a" and "an", expressed as three shares. In a ScriptGrain voice profile it's called the_a_an_ratio, sits under Function words, and gets read from your samples by the extraction model rather than set by you. It feeds into voice match scoring with a weight of 0.75.

What is the a an ratio?

It's the relative share of "the", "a" and "an" in a piece of writing, normalised so the three shares add to one. It's described in the app as an unconscious stylistic marker: something writers do without meaning to, which is what makes it a useful signal for telling one writer's habits apart from another's.

How is article balance scored?

Article balance is scored as part of voice match, with a weight of 0.75. Code counts "the", "a" and "an" in your draft, normalises them to shares, then compares those shares against your profile's stored shares. Similarity is one minus half the summed absolute difference between the two sets. Closer shares mean a higher score on this component.

How are articles used in stylometry?

Article choice is treated as a function word feature: low-level, largely unconscious, and consistent across a writer's output. Because writers don't deliberately control their "the" versus "a"/"an" balance, it tends to hold steadier across topics and genres, which can make it more reliable as a fingerprint than word choices you make on purpose. ScriptGrain groups it with other function words for this reason.

How do I make AI match my writing habits for articles?

Give it real samples of your own writing rather than instructions about ratios; nobody can accurately state their own "the" versus "a" balance from memory. A ScriptGrain profile reads it from your samples along with the other 44 attributes, then applies it as a hard constraint whenever it generates text, whether that's through the API, the MCP server, or the app itself.

Related

Sources