Pronoun analyser

By Jack Stovell · published 2026-09-24

A pronoun analyser measures how a text's I, we and you pronouns split between the three, as shares that add up to 100%. ScriptGrain stores this as pronoun_distribution, one of 45 voice attributes, and weights it 1.5 in the voice match score. In ScriptGrain's reference corpus of 299 public pieces, the median I-share is 35%.

Analyse your pronoun distribution free

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

Try the six-signal check first

At a glance

What pronoun distribution measures

Pronoun distribution measures how often you write from your own point of view, how often you speak for a group, and how often you turn to face the reader directly. The app's own definition is blunt: "the share of I / we / you pronouns." No tone, no sentiment: just the mix of three sets of function words.

When a draft is scored, code counts I-forms (I, me, my, mine, I'm, I've, I'll, I'd), we-forms (we, us, our, ours, we're, we've, we'll) and you-forms (you, your, yours, you're, you've, you'll), then turns each into a share of the three. Write "I think you'll like this" and you've used one I-form and one you-form; the count doesn't care what you think, only that you said "I" and then turned to address someone.

The free habits report counts the same forms in code and turns them into readable lines: 'Talking to the reader' for your you-share, 'First person' for your I-share. At or above the 75th corpus percentile it names the habit: 'You address the reader directly' or 'First person throughout'. The value stored in your voice profile is a different thing: the extraction model's reading of your samples, with Claude Sonnet 5 reading each sample in pass one and synthesising the profile in pass two. Code counts pronouns only in a draft being scored or in the free tools.

How pronoun distribution is scored

Pronoun distribution is scored in the voice match score, with a weight of 1.5. Both your draft and your profile get reduced to three shares (I, we, you), and similarity is calculated as one minus half the summed absolute difference between them. Write mostly in "I" when your profile is built on "you", and the gap shows up straight away.

Code counts the draft's pronouns directly, so the draft side of the comparison is arithmetic; the profile side is the extraction model's reading. Twenty-eight of the 45 attributes can move your voice match score; this is one of them, and a meaningful one given the 1.5 weight. For the full mechanics of how features like this combine into a single 0 to 1 score, with Claude Haiku judging softer things like rhythm and formality alongside code-counted features like this one, see the voice measurement framework and the voice match glossary entry.

How to analyse pronoun distribution in your own writing

Analysing your own pronoun distribution by hand takes about ten minutes and a highlighter, or your editor's find function.

  1. Take a writing sample of at least a few hundred words, ideally something typical of your usual output rather than a one-off.
  2. Search and count every I-form: I, me, my, mine, I'm, I've, I'll, I'd.
  3. Search and count every we-form: we, us, our, ours, we're, we've, we'll.
  4. Search and count every you-form: you, your, yours, you're, you've, you'll.
  5. Add the three totals together, then divide each individual count by that total and multiply by 100 to get your three shares.
  6. Compare the shares. A heavy I skew reads as personal essay or memoir. A heavy you skew reads as instructional or sales copy. Balanced or we-heavy tends to read as collaborative or corporate.

Do this across a few samples and you'll start to see your own default. That's your baseline, and it's worth knowing before you try to change it.

ScriptGrain's free voice profile reads this attribute, and the other 44, from a writing sample.

Typical pronoun distribution in published writing

In published writing, I-forms have the highest median share of the three, according to ScriptGrain's reference corpus: 299 public pieces from 47 sources (835,655 words), built 2026-09-20, measured in code and then discarded. Across that corpus, I-forms sit at p10 3, p25 13, median 35, p75 56, p90 73 (as a percentage of I/we/you pronouns). We-forms are lower across the board: p10 1, p25 6, median 19, p75 44, p90 69. You-forms have the widest spread and the highest p90: p10 4, p25 14, median 28, p75 51, p90 83.

Read plainly: the typical piece leans slightly towards "I", "we" has the lowest median of the three, and a quarter of the corpus gives 51% or more of its pronouns to "you". If your writing sits near the p90 you-share of 83%, you're writing squarely in "talk to the reader" mode. Sit near the p90 I-share of 73% and you're deep in personal essay territory.

How to make AI writing more personal

Making AI writing more personal means deliberately shifting the pronoun mix toward I and you, rather than hoping the model guesses your intent.

  1. Write from your own point of view. Use "I" wherever you are genuinely the source of an opinion or experience.
  2. Address the reader directly as "you" whenever you're giving advice or instructions.
  3. Reserve "we" for what you and a team, or you and the reader, actually share; don't reach for it as filler.

A ScriptGrain profile sends your stored pronoun shares in the voice profile JSON of every generation prompt as a hard constraint, and the conversational register adds its own line: "more direct address ("you", "we")". The profile reaches ChatGPT, Claude and other tools through the API and MCP server. Its siblings article use and discourse markers are small-word habits of the same kind.

How to make AI writing less personal

Making AI writing less personal means stripping out first-person asides and reserving "you" for genuine instructions.

  1. Ask explicitly for the third person about the subject, rather than the first person about yourself.
  2. Remove first-person asides; cut the "I think" and "in my experience" lines even where they feel natural.
  3. Keep "you" only where you're giving direct instructions, not as a general conversational habit.

The same mechanism works in reverse: a profile built from impersonal samples sends its pronoun shares as a hard constraint, and when a draft is scored, code checks the draft's shares against them. The profile reaches other tools through the same API and MCP server. The rest of the function words group, including paragraph opener words, works on the same small-word level.

Questions

What is a pronoun analyser?

A pronoun analyser counts how often a piece of writing uses I, we and you, then expresses each as a share of the total. In ScriptGrain, this is the pronoun_distribution attribute, one of 45 tracked in a voice profile under the function words group, stored as three shares, one each for I, we and you.

First person vs second person writing: which should I use?

There's no universal answer; it depends on what you write. In ScriptGrain's reference corpus of 299 public pieces the median I-share is 35% and the median you-share 28%, so both are common. First person suits opinion and memoir; second person suits instructions and sales copy. Check your own baseline first (see the analysis steps above), then decide deliberately rather than by habit.

How is pronoun use scored in a voice match?

Pronoun distribution is scored as part of voice match with a weight of 1.5. Code counts your draft's I, we and you shares directly, compares them to your profile's shares, and calculates similarity as one minus half the summed absolute difference. See the voice measurement framework for how this combines with other attributes.

How do I make AI write in first person?

Ask explicitly for "I" statements where you're the source of an opinion or experience, rather than a neutral third-person voice. A ScriptGrain profile stores your natural I-share and sends it as a hard constraint in the system prompt, so its drafts aim at your usual first-person habits instead of a generic, distant voice.

How do I make ChatGPT talk to the reader?

Address the reader directly as "you" in instructional or advisory sentences, and reserve "we" for genuinely shared ground. A ScriptGrain profile carries your stored you-share into the system prompt as a constraint, and reaches ChatGPT, Claude and other tools through the API and MCP server, so connected tools read the same habit instead of starting from scratch each session.

Related

Sources