ScriptGrain vs ChatGPT

ScriptGrain makes ChatGPT better at being you. Custom instructions apply what you can put into words about yourself. ScriptGrain measures your voice as 45 attributes from writing you have already published, connects to ChatGPT so it writes from that profile, and scores every draft against the measurement.

The actual difference

Custom instructions rely on what you can articulate about your own style, and most of a voice is habits you cannot see: function-word patterns, punctuation cadence, clause ordering. ScriptGrain measures 45 attributes from writing you have already published, including those unconscious markers, then plugs into ChatGPT through its connector so the assistant writes from the measured profile and every draft is scored against it. It is not a rival to ChatGPT; it is the part ChatGPT does not have.

Custom instructions cost nothing on any ChatGPT plan, and the paid tiers buy a better model and higher limits rather than a better memory of your voice. ScriptGrain's Free plan measures your voice with no card; connecting it to ChatGPT is included on every plan, and the paid plans (from £12/month) add regular generation, polishing and the monthly voice monitor.

ScriptGrain is not a replacement for the assistant you already use. It measures your voice from writing you have published and hands that profile to ChatGPT, Claude, Muse or any MCP client through its connector, then scores every draft against it. ScriptGrain makes your AI better at being you.

Connect ScriptGrain to ChatGPT

Every AI voice tool compared: prices, scores and API access, checked

	ScriptGrain	ChatGPT
Entry price	£12/mo (Writer, 1 voice profile)	Free (custom instructions are on every plan)
Free tier	Yes, See your own voice profile and try one draft. No card.	ChatGPT's free tier includes custom instructions
Voices on entry plan	1 (3 on Operator, £29/mo)	One set of custom instructions per account
How the voice is built	45 measured style attributes, extracted from writing you have already published	Custom instructions: free text you write about yourself and how you want replies, up to 1,500 characters on Free and Go and 5,000 on paid plans, applied to every chat. Projects can add example files and instructions of their own. It applies what you can articulate about your style.

ChatGPT figures read from their pricing page on 9 August 2026.

What ChatGPT is built for

ChatGPT is OpenAI's general assistant. One product, many jobs: research, code, images, browsing, writing. That breadth is the point, and it's a genuine strength. Nobody opens it to get a narrow thing done well. They open it because it does most things well enough, in one place.

Personalisation sits on top of that. Custom instructions are the layer applied to every chat, so the assistant doesn't start cold each time. Custom GPTs and projects extend the idea: each carries its own instructions and files, which gives you free per-context setups. A GPT for support replies, a project for a book, that sort of thing.

ChatGPT is built to be useful to everyone. A tool built for everyone has to treat your voice as one more preference among hundreds. That's fair enough. It isn't a flaw, just a design choice, and it explains everything in the next section.

How ChatGPT handles voice

There are three mechanisms, and they work differently.

Custom instructions are two free-text boxes: what ChatGPT should know about you, and how it should respond. The limits are 1,500 characters each on Free and Go, and 5,000 characters each on Plus, Pro, Business, Enterprise and Education. Everything you want to say about your voice has to fit in there. That's a short essay at best.

Memory is a separate feature. It saves details you mention or ask it to remember and applies them to later chats. You can view it, delete it, or turn it off. Useful for facts about you (your job, your audience, your preferences). Less obviously useful for cadence.

Custom GPTs carry their own instructions, up to 8,000 characters, plus uploaded knowledge files. Drop in a few samples of your writing and the model can draw on them. That's a real technique, and it costs nothing extra.

Now the catch. All three are free text. None is a measured profile, and nothing scores a draft against what you wrote. You describe your voice, the model interprets the description, and you judge the result by eye. If the draft feels off, you tweak the wording and try again.

And to be fair, that loop works when your voice is easy to describe. "Short sentences, dry humour, no exclamation marks" is a workable brief. The trouble is that most of a voice isn't in that list. It lives in habits you can't see from the inside: how often you reach for a comma, whether you put the conclusion first, which little words open your paragraphs. You can't write down what you've never noticed. So the instructions capture the part of your voice you can articulate, and the model fills the rest with its own defaults.

How ScriptGrain handles voice

ScriptGrain starts from the other end. You don't describe your voice; you hand over writing you've already published, and it measures that.

The result is a profile of 45 attributes in 8 layers: lexical, syntactic, tone and register, rhetorical, punctuation and format, function words, content patterns, and quirks and cadence. The attributes are concrete. Average sentence length and its variance. Contraction rate. Comma density. Semicolon and dash habits. Formality on a 0 to 10 scale. Confidence versus hedging. Opening style. Argument structure (evidence-first, conclusion-first, narrative). Paragraph openers. Pronoun mix. The attribute reference lists them all, and the measurement framework publishes the method.

You need at least one sample of 50+ words. Three or more pieces and about 3,000 words is the recommendation. Extraction usually takes under a minute; the median was 39 seconds when measured on 2026-09-28.

Then the part that matters most. Every draft gets a voice match score from 0 to 1. On-voice drafts typically land between 0.85 and 0.95. Partial matches sit between 0.5 and 0.8. A clearly different voice comes in under 0.45. The score arrives with per-feature deltas, so you see the draft's value against the profile's value for each attribute. Skip "this feels stiff" and say "your contraction rate is half the profile's."

Polish takes a draft and revises it toward a target score (the default is 0.9). Drafts are written with the profile as a constraint, and the app learns from the edits you save.

The profile reaches ChatGPT, Claude, Gemini, Cursor, Meta Muse and any MCP client through the MCP server and the REST API. It's read live each time, and both routes are on every plan including Free. It is never a pasted prompt. If you want the plumbing, see the API and MCP page.

The same job in each tool

Take one job: a 900-word founder post about why your team dropped a feature. You write punchy, open with the conclusion, use contractions freely, and keep commas sparse.

In ChatGPT alone, you'd write your custom instructions first. Something like: direct, conversational, short sentences, conclusion first, dry. Maybe you paste a sample into a custom GPT. You ask for the post. What comes back is competent and recognisably in the neighbourhood. The opening is probably a hook rather than the conclusion, because the model's default leans that way and "conclusion first" was one phrase among several. Paragraphs run a little even. There's a tidy sign-off you'd never write.

So you edit. You cut the sign-off, break up the rhythm, flip the opening. Next time you add a line to the instructions about sign-offs. The box fills up. By the third or fourth post you're maintaining a document of corrections, and you're the only quality check in the loop.

With ScriptGrain connected, the job starts differently. Your profile already holds the conclusion-first habit, the contraction rate, the comma density and the sentence-length variance, all measured from things you published. ChatGPT writes the post reading that profile live. The draft comes back with a voice match score and the deltas. Say it lands in the partial band: the deltas show the sentence lengths are too uniform and the opener is wrong. Polish runs toward the 0.9 target and you review the result.

You still read it. You still edit. But the feedback is specific, repeatable and on the page, instead of a hunch about what's off.

Which one fits which situation

Pick ChatGPT on its own if:

You write occasionally, and a rough match to your style is good enough.
Your voice is easy to describe in a paragraph or two, and you're happy to judge drafts by eye.
Most of what you do with the assistant isn't writing at all: research, code, images, browsing.
You want free, per-context setups through custom GPTs and projects.

Pick ScriptGrain if:

Writing in your voice is a recurring job, and drift between pieces costs you something.
Several people write under one brand and need a shared, measurable standard.
You can't put your style into words, but you can point at twenty things you've published.
You want a number and a breakdown rather than a feeling. Voice drift is hard to catch when the only instrument is your own ear.

Use both when you already live in ChatGPT and like it. That's most people reading this page. You keep the assistant, the model and the breadth, and add the measurement it doesn't have. It isn't a rival setup; it's a layer.

One fair note on cost. Custom instructions are free on every plan, and ChatGPT's paid tiers (as read on 2026-09-30: Go $8 a month in the US, Plus $20, Pro $200) buy a better model and higher limits, not a better memory of your voice. ScriptGrain's Free plan measures one profile with no card, and the pricing page has the paid plans from £12 a month.

Using ScriptGrain alongside ChatGPT

Nothing gets migrated. Your custom instructions, memory, projects and custom GPTs stay exactly where they are. The connection is additive, and the setup is covered on the ChatGPT integration page. ChatGPT connects to remote MCP servers as a client through connectors, though availability varies by plan, so check yours before planning around it. The OpenAI API is a separate, metered product; that's worth knowing if you're wiring things up programmatically.

A sensible order for a first week:

Collect three or more pieces you're proud of, roughly 3,000 words in total.
Build a free profile from them and read it. Some of it will surprise you.
Connect it to ChatGPT and write one real piece, not a test.
Read the score and the deltas before you read the draft.

Once that works, trim your custom instructions. Keep the facts (role, audience, topics) and drop the style guesswork, since the profile now carries that. You'll free up characters, which is handy on the 1,500 limit.

Two more checks are worth having nearby. The brand voice consistency checker is useful when several writers feed one profile, and the AI cliché checker catches the stock phrases that creep into any assistant's output. If you use Claude as well, the same profile travels there too; see the Claude page.

Checking your voice is one half of it. The other half is how assistants describe you when nobody's prompting them. Ask one who you are and you'll often get something technically true and completely flat, a description that fits half of LinkedIn. We've written up why assistants get your brand wrong, what that flattening costs you, and what you can do to fix it.

If you want a wider view first, the tools matrix and the comparison index put this one beside the rest.

A last practical point. A profile is a measurement of writing you've already published, so it ages as you change. Re-run it when your style moves on. Most people skip that step, and it shows.

Where ChatGPT is stronger

The default AI writing surface for most people. Custom instructions are its personalisation layer: free text about you and your preferences, applied to every chat.

A full general assistant: research, code, images, browsing and writing in one place
Custom instructions are free on every plan and take two minutes to set up
Excellent at following explicit instructions once you can state them

If you want one assistant for research, code, brainstorming and quick drafts, ChatGPT is that product, and for casual replies where voice does not matter, custom instructions are plenty. Keep ChatGPT; give it your measured voice.

Questions

Is ScriptGrain better than ChatGPT?

They solve the problem differently. Custom instructions rely on what you can articulate about your own style, and most of a voice is habits you cannot see: function-word patterns, punctuation cadence, clause ordering. ScriptGrain measures 45 attributes from writing you have already published, including those unconscious markers, then plugs into ChatGPT through its connector so the assistant writes from the measured profile and every draft is scored against it. It is not a rival to ChatGPT; it is the part ChatGPT does not have. If you want one assistant for research, code, brainstorming and quick drafts, ChatGPT is that product, and for casual replies where voice does not matter, custom instructions are plenty. Keep ChatGPT; give it your measured voice.

How much does ChatGPT cost compared with ScriptGrain?

ChatGPT's entry plan is Free (custom instructions are on every plan), with this free tier: ChatGPT's free tier includes custom instructions. ScriptGrain starts at £12/month (Writer, 1 voice profile) with a free plan: See your own voice profile and try one draft. No card. ChatGPT figures read from their pricing page on 9 August 2026.

Does ChatGPT learn my writing voice?

Custom instructions: free text you write about yourself and how you want replies, up to 1,500 characters on Free and Go and 5,000 on paid plans, applied to every chat. Projects can add example files and instructions of their own. It applies what you can articulate about your style. ScriptGrain instead measures 45 style attributes from writing you have already published, and scores every draft against that measured profile.