An AI writing tool that sounds like you, with a number to prove it

Most AI writing tools ask you to describe your voice in adjectives. ScriptGrain measures it from writing you have already published: 45 attributes, from sentence rhythm to the words you reach for without noticing. Every draft then comes back with a voice-match score against that profile, so you can see how close it got instead of guessing.

Why a score matters

"Sounds like me" is not a setting you can check. A contraction rate, a comma density and a sentence-length variance are. ScriptGrain measures them in your samples, holds every draft to them, and reports the distance as a number. When a draft falls short, a polish loop revises toward your profile and re-scores, so the writing moves toward you rather than toward the model's habits.

What is set up for the US

What a US business should look for

Start with the boring stuff. It's the stuff that bites you later.

First, American English has to be a real setting, not a cleanup pass you run afterward. In ScriptGrain it's applied at generation, so "organization" and "color" show up as written instead of being patched in. And a profile keeps the spelling its samples use, so if your team writes "program" and "center," the drafts do too.

Second, pricing. You want dollars, and you want them before tax. Writer is $17 a month for one voice profile. Operator is $40 a month for three. Sales tax gets added at checkout only where it applies, so the number you see is the number you compare. There's also a free plan, and it needs no card.

Third, where your data goes. More on that below, because it deserves its own section.

Fourth, and this is the one most buyers skip: do the drafts sound like you? Not "professional." Not "friendly." Like your business, specifically. Every tool claims it does this. Only some can show you a number. ScriptGrain scores every draft from 0 to 1 against your profile. On-voice drafts typically land between 0.85 and 0.95. Partial matches run 0.5 to 0.8. Anything under 0.45 is clearly a different voice. So you're never guessing whether the draft is close; you're reading a score.

And to be fair, plenty of general-purpose assistants write clean, useful copy. If you only need a quick first pass and don't care whose voice it's in, that may be enough. The gap opens when three people send customers emails that sound like three different companies.

American English goes deeper than spelling

Spelling is the easy layer. Anyone can swap "colour" for "color."

The harder part is register and habit. Consider two sentences a US contractor might send a client:

  1. "We'll have the estimate over to you by Friday; let us know if the scope changes."
  2. "Please find attached our estimate for your consideration. Should the scope be amended, do advise."

Both are correct English. Only one sounds like someone who'd show up on time. The second reads like a translation of the first into a suit.

That gap lives in things a spell checker can't see. Contraction rate. Average sentence length and how much it swings. Comma density. How often you reach for a semicolon or a dash. Whether you open with the point or build to it. Formality on a scale of 0 to 10. How much you hedge versus just say the thing.

ScriptGrain measures those as part of a 45-attribute profile across 8 layers: lexical, syntactic, tone and register, rhetorical, punctuation and format, function words, content patterns, and quirks and cadence. It builds the profile from writing you've already published, not from a questionnaire about how you think you write. People are bad at describing their own voice. Their samples aren't.

So a US business gets two things at once. The spelling is right, and the habits are yours. One without the other gives you copy that's correct and still wrong.

Where your writing goes

Plain account, no fog.

ScriptGrain is built and run by Adapt Progress Evolve Limited, a UK company (England and Wales, company 10047328). Account data is stored in the EU (Ireland) with Supabase. Two US providers touch your writing. The Claude models from Anthropic measure a voice and draft in it. Voyage AI makes the embeddings used in scoring.

Because the company is UK-based, data can leave the UK. When it does, the provider signs the UK International Data Transfer Agreement or the UK Addendum to the EU Standard Contractual Clauses. That's the paperwork that makes the transfer lawful, and it's worth knowing it exists if your legal person asks.

Now the list of things that never happen:

You stay in control of the rest. You can download everything we hold, or delete your account and its samples immediately, from Settings. API keys are stored only as a hash, so even we can't read one back to you.

Fair enough if you want to read the source yourself. The details sit on the trust page and in the privacy policy, both checked on 2026-09-30.

Starting small

Don't buy anything first. Use the free plan: one profile, one full analysis, no card.

Pick your samples. A profile needs at least one sample of 50 or more words, so you can technically start with a single email. Better is three or more pieces and about 3,000 words. Choose writing you'd be happy to have copied: the newsletter that got replies, the proposal that won the job, the About page you didn't hate. Extraction usually takes under a minute; the median was 39 seconds when we measured on 2026-09-28.

Then read the result before you write anything. Does the profile describe you? If the formality score looks stuffier than you think you are, your samples probably are. That's useful to learn.

From there, the path is short. A solo owner or a one-person marketing shop fits on Writer: one profile, $17 a month. A small team with a few distinct voices, say the company itself, the founder, and a support lead, fits on Operator: three profiles, $40 a month.

One more thing that matters more than it sounds. Every plan includes the API and the MCP server, Free included. The profile reaches ChatGPT, Claude, Gemini, Cursor, Meta Muse and any MCP client, and it's read live each time. It's never a pasted prompt that goes stale in someone's notes app. So the tools your team already uses can write in your voice without anyone changing their routine.

And when a draft comes back at 0.78? Polish revises it toward a target score, 0.9 by default. The app also learns from the edits you save. So the second month tends to go better than the first. That's the whole idea.

See your own voice profile free

The free plan analyzes your writing across all 45 reference points, no card required. Generating in that voice is the paid step. US pricing: scriptgrain.com/us/pricing. Writing in British English? See the UK edition of this page.