ChatGPT could soon write in your voice, here’s how to use that safely

ChatGPT is moving towards matching your writing style, learning how you phrase things by reading your Gmail, Slack and Docs, so it can draft emails and proposals that sound like you. That saves real time, but it only works safely if you review every output before it goes out and you decide, up front, what data you actually let it touch. This post covers how to test the feature on low-risk drafting and how to set clear rules so you keep control.

The problem with a ChatGPT writing style that sounds like you

Sounding like yourself is the whole point. It’s also the risk. When ChatGPT drafts an email that reads exactly the way you write, you stop reading it as carefully. You skim, you nod, you hit send. That’s when a wrong figure, a made-up detail or the wrong tone for a difficult client slips through, wearing your voice.

The second problem sits behind the scenes. To copy your style, the tool needs examples of your writing. That means connecting it to the places your writing lives, your inbox, your chat, your documents. Those places also hold client names, pricing, contracts and private conversations. Convenience and exposure come through the same door.

What it actually costs you

The time saving is genuine. Drafting a proposal from a blank page is slow, and having a first draft in your own voice cuts that down a lot. But the cost of getting it wrong is not just an awkward email. It’s a client who reads something that isn’t quite true, or data that ends up somewhere you didn’t intend, or a habit of trusting output you never checked.

Most owners I speak to don’t have a data problem because they’re careless. They have one because nobody stopped to decide what should and shouldn’t be connected before switching the feature on. The default is usually “connect everything,” and the default is rarely what you’d choose on purpose.

How to test it without getting burned

Having run my own businesses for 19 years and worked with tools like HubSpot, n8n and Claude, my approach is simple. Test on the boring stuff first, prove it, then decide what to trust it with.

Start with low-stakes drafts

Point it at internal emails, follow-ups and first drafts of proposals. Not the final client-facing send. Let it learn your phrasing on messages where a mistake costs nothing. You’ll quickly see where it nails your tone and where it overreaches.

Review every output before it goes out

Treat every draft as a draft. Read it as if a new junior wrote it in your name, because that’s effectively what’s happening. Check the facts, the figures, the client details and the tone for that specific relationship. The style will be right long before the substance is.

Decide what data you connect, on purpose

Before you link your Gmail, Slack or Docs, decide what’s in and what’s out. Do you want it reading client contract threads? Pricing discussions? Personal messages? For most owners the answer is a small, deliberate set of sources, not everything. OpenAI’s own help documentation and its privacy policy are worth reading before you connect anything, so you know how your data is handled.

Write down the rules

If anyone else in your business uses these tools, the rules can’t live in your head. A single page covering what’s connected, what always gets reviewed and what never goes near the tool is enough. The UK’s ICO guidance on AI is a sensible reference point if you handle client or personal data.

Why the review step is non-negotiable

The clients we’ve worked with for years get value from automation because it removes the manual grind, not because it removes their judgement. As Swann Virassamy at Cromebloc put it, “Luke is a problem-solver. He rebuilt our onboarding and suggested optimisations we had not considered. It streamlined our client onboarding and saved us a huge amount of manual work.” The saving comes from taking the repetitive weight off, while a person still owns the final call. A style-matching tool is exactly the same. Let it do the first draft. You keep the decision.

What to do tomorrow

Turn the feature on for one thing, drafting a proposal or a batch of follow-up emails. Set one rule, nothing goes out unread. And write down which of your accounts you’re happy to connect and which you’re not. That’s it. You’ll know within a week whether it’s saving you real time or just making you feel busy.

If you want to build these habits properly and see how tools like this fit into your day-to-day, join our next AI workshop. We’ll walk through using AI safely on real business tasks, without the guesswork. That’s the kind of practical grounding 07 Heaven Systems is built to give you.

Common questions

Can ChatGPT copy my writing style?

Yes, ChatGPT is moving towards matching your writing style by learning from examples of how you write, including content in connected accounts like Gmail, Slack and Docs. The output can sound convincingly like you, which is useful for drafting, but it means you need to review everything before sending and control which accounts you connect.

Is it safe to connect ChatGPT to my Gmail and Docs?

It can be, if you do it deliberately. The risk is connecting everything by default, including threads with client data, pricing or private conversations. Decide which sources are in and which are out before you link anything, read the provider’s privacy policy, and if you handle personal data check the ICO’s guidance on AI first.

Should I let ChatGPT send emails for me automatically?

No, not without a review step. The tool can draft in your voice, but it can also get facts, figures and tone wrong while sounding exactly like you. Treat every output as a first draft, read it as if someone else wrote it in your name, and keep the final decision to send with a human.

Rather have it built for you? See what an AI automation agency actually does and how we build and run these systems for UK businesses.