Analyse a LinkedIn writer's style: free stylometric report from public posts

By Jack Stovell · published 2026-09-24 · updated 2026-09-20

Paste three or more of a LinkedIn writer's posts (120+ words total) and the report measures the habits that make them recognisable: sentence rhythm, contractions, punctuation, pronoun mix, word length, and AI-tell density. It runs free, in the browser, and stores nothing.

Here's the thing about LinkedIn posts: they all sort of look the same until you measure them. Paste yours below, a client's (with agreement), or a public figure's for study, and the report renders underneath this paragraph. No login. No waiting.

What the report measures

So what actually gets counted. Every piece runs through the same set of features, the counted ones and the judged ones. Counted: sentence length and variance, contractions, commas, exclamations, semicolons, brackets, questions, ellipses, pronoun mix, article balance, word length, signature phrases. Judged: formality, humour, expressiveness, confidence, opening style, argument structure, rhythm, complexity, clause ordering, specificity.

Those get weighted and calibrated to a score between 0 and 1. Against a saved profile, 0.85 to 0.95 reads as a genuine voice match, 0.5 to 0.8 is partial, under 0.45 is off. Without a profile, comparing two pasted pieces instead, the bar sits differently: 0.80 and above counts as voice, 0.55 to 0.79 is drifting, under 0.55 is off. The free check needs 120 words per piece and keeps none of it. Full method, if you want the maths, lives at the measurement framework.

One limitation worth stating plainly: LinkedIn pages can't be fetched by URL. So you paste. That's not a workaround, it's the only path in.

Reading a LinkedIn writer's habits

You know the shape of a LinkedIn post before you've read a word of it. One-line paragraphs. A question near the top. A short sentence doing the work three long ones couldn't. That's not accidental, it's a rhythm, and it's exactly what the counted features pick up: the share of short sentences against the rest, the variance between a five-word line and a thirty-word one sitting next to it.

Contractions per 1,000 words tell you how conversational someone actually is, versus how conversational they think they sound. Commas per sentence tell you whether someone builds long, layered thoughts or fires off fragments. Questions in the body (yes or no, simple as that) tell you whether someone's performing engagement or genuinely asking something.

None of this needs a stats degree to read. If a writer's sentence variance is high and their short-sentence share is high too, you're looking at someone who punches. If their pronoun mix leans heavily "you", they're writing at the reader, not about themselves. The report doesn't interpret these for you with adjectives; it places each measurement against a reference corpus of 299 public pieces and gives you a percentile. Where does this writer's comma use sit against everyone else's? That's the question it answers.

Attribution and the line we hold

Studying a voice and impersonating a person are different things, and the difference matters. You can run the habits report on your own posts to see your patterns clearly. You can run it on a client's posts, with their agreement, before writing in their voice for them. You can run it on a public figure's posts to study a style that's taught you something.

What you can't do: use it, or anything downstream of it, to impersonate a named living person. ScriptGrain doesn't support that, full stop. The report measures pattern, not permission. It's on you to hold the line on what the pattern gets used for afterwards.

To be fair, this isn't a moral lecture bolted onto a tool page. It's a practical boundary, because a voice profile is genuinely powerful once you're generating from it, and powerful things need a stated limit.

From a report to a profile

The free habits report gives you a one-off read: paste, measure, done. A saved profile is different. It's 45 attributes across eight layers (lexical, syntactic, tone and register, rhetorical, punctuation and format, function words, content patterns, quirks and cadence), extracted in two passes: one that reads each sample separately, one that synthesises all of them into a single profile. You'll need at least one sample of 50+ words, though the product recommends three or more pieces and around 3,000 words for anything reliable. Extraction takes roughly one to three minutes, and every profile comes with a confidence score between 0 and 1, so you know how much to trust it.

A free account gets you one profile, held to those 45 attributes; the full list lives at the writing voice attributes reference. Compare that to Taplio's style matcher, which (per its site, checked 20 September 2026) sits behind a subscription, works LinkedIn-only, and gives no attribute-level breakdown at all. You get a score, not a why.

Once a profile exists, it travels. It's what powers writing LinkedIn posts in a voice, what the broader style analysis tool measures the same way, and what sits underneath AI style mimicry and the LinkedIn post generator use case. The report is the starting point. The profile is what it's for.

Questions

Does the tool fetch posts from a LinkedIn URL?

No. LinkedIn pages can't be fetched by URL, so the report only works on pasted text. Copy three or more posts, 120 words or more in total, and paste them in. That's the only input path, and it's deliberate rather than a missing feature.

How much text do I need for a useful reading?

The free habits report needs 120+ words per piece to run at all. For a full 45-attribute profile, one sample of 50+ words is the technical minimum, but the product recommends three or more pieces totalling around 3,000 words for a reading you can actually trust.

Can I analyse someone else's posts without their permission?

You can study a public figure's posts for that purpose alone. You can analyse a client's posts with their agreement. What's not supported, under any circumstance, is using the tool or its outputs to impersonate a named living person. Study is fine; impersonation isn't.

What's the difference between the free check and a saved profile?

The free check compares two pasted pieces on the spot and stores nothing. A saved profile extracts all 45 attributes across eight layers from your samples, keeps a confidence score, and travels by API and MCP into other tools, like the post generator or the style mimic, so you're not re-pasting every time.

What counts as a good voice-match score?

Against a saved profile, 0.85 to 0.95 counts as a genuine match, 0.5 to 0.8 is partial, under 0.45 is off. Comparing two pasted pieces with no profile, 0.80 and above counts as voice, 0.55 to 0.79 is drifting, under 0.55 is off.