How to Write with AI Without Losing Your Voice

AI can help you write faster. Most people have worked that out by now.

The bigger question is whether it helps you write better. And perhaps more importantly, whether the final piece still sounds like you.

Because that’s where things often start to go wrong. You ask AI to help with an email, proposal, report or LinkedIn post. It gives you something polished, tidy and grammatically perfect. But when you read it back, something feels slightly off. It sounds professional and competent… it just doesn’t sound like you.

The risk is not just that AI writes badly. The risk is that it writes acceptably, and we stop noticing whether it is useful.
So, how do you write with AI without losing your voice?

The answer is surprisingly simple: use AI as a thinking partner, not a ghost writer. The aim is not just to make writing easier for you. It is to make reading easier for the person on the other end.

Most advice about writing with AI focuses on prompts. I think that’s the wrong place to start. Good writing doesn’t begin with writing. It begins with thinking. And AI is often most valuable before the first draft ever exists.

Get AI to Help You Think Before You Get It to Write

To write with AI without losing your voice, use AI to help you think before you ask it to write. Explore ideas, challenge assumptions, clarify your message and gather your thoughts first. Then brief AI properly and edit the final output yourself. The goal isn’t to sound polished and generic. The goal is to sound clear, useful and recognisably human.

Writing with AI means using AI to help you think, structure, edit and refine your writing. It does not have to mean asking AI to create the whole piece from scratch. The best results usually come when human judgement stays in control.

Most People Start at the Wrong End

One of the most common pieces of AI advice is: “Get AI to write the first draft.”

Sometimes that’s fine. But for important communication, “write me a first draft” is often the laziest prompt, not the smartest one. And if you’re trying to preserve your voice, it’s usually the wrong place to start.

I think writing with AI is a bit like decorating a room. People with little decorating experience get the paint out within half an hour. Experienced decorators know that most of the work happens before the paint ever touches the wall. They clean, fill, sand, prepare and prime. The painting itself is often the quickest part.

Writing works in much the same way. The visible writing is only the final stage. The real work happens beforehand.

When I’m writing a proposal, report or blog article, I rarely jump straight to a draft. I’ll usually spend time downloading my thoughts into AI, sharing scraps of ideas, uploading meeting transcripts, challenging my assumptions, exploring different angles, testing arguments and identifying gaps in my thinking.

By the time we get to creating a first draft, AI and I are already aligned on what we’re trying to say.

The draft isn’t the starting point. It’s the outcome of all the thinking that came before it.

That’s one of the biggest differences between using AI as a writing tool and using AI as a thinking partner. The outputs are sharper, briefer, more natural and actually sound like the human behind them.

The Biggest Writing Problem AI Is Creating

Most people assume AI’s biggest benefit is helping us write more quickly. And it does.

But in my own use of AI, and in conversations with clients and colleagues, I’ve noticed something else happening: AI often makes people write more. Not better or clearer. Just more.

One colleague’s client emails gradually became longer and longer after she started using AI regularly. The emails weren’t bad. In fact, they were full of useful information. The problem was that they contained too much of it. Because AI made writing easier, she could include every detail, every explanation and every caveat without feeling the effort of writing it. The writer’s workload disappeared. The reader’s workload didn’t.

That’s when I realised something interesting. We’re creating more content because we can, not because we should. And the result can be absurd.

Someone receives a long AI-assisted email and thinks, “That’s too long. I’ll get AI to summarise it.” They then use AI to write the reply. The sender receives that reply and uses AI to summarise it… Before long, two AI tools are effectively having an email conversation while the humans are barely involved.

That’s not better communication. It’s just more words moving around.

Writing with AI shouldn’t remove humans from the loop. It should make the human contribution more valuable.

A Simplified Example: When AI Makes an Email Too Long

Here is a simplified example of the pattern. It is not a real client email, but it reflects something I have seen happen: AI helps someone include everything, when the reader only needs the essentials.

Before: AI-Assisted, but Too Long

Subject: Update on the project timeline

Hi Sam,

I hope you’re well.

I just wanted to provide a quick update on the project timeline following some internal conversations this week. As you know, we’ve been working through the final stages of the planning process and have been reviewing the different components involved to make sure everything is aligned before we move into the next phase.

There have been a few moving parts, including some changes to availability across the team, a couple of points that required further clarification and some additional checks we wanted to complete before confirming the final dates. We are still confident that everything is moving in the right direction, but we now think it would be sensible to allow an additional week before sharing the final version of the plan.

This means we are looking at sending the updated plan next Friday rather than this Friday. We don’t believe this will affect the overall delivery window, and we will continue to monitor the timeline closely to make sure everything stays on track. We really appreciate your patience while we work through the final details.

Please let me know if you have any questions or if there is anything else you would like us to consider at this stage.

Best wishes,

After: Shorter, Clearer and More Useful

Subject: Updated project plan coming next Friday

Hi Sam,

A quick update on the project timeline.

We need one extra week to finalise the plan, mainly because we’re checking team availability and a couple of details before we confirm dates.

I’ll send the updated plan next Friday. At this stage, we don’t expect this to affect the overall delivery window.

No action needed from you for now, but I’ll flag anything that needs a decision.

Best wishes,

The second version is not cleverer. It is just kinder to the reader.

It gets to the point, explains the reason, gives the next step and removes the padding.

That is what human judgement adds. AI can help create the first version, but a person still needs to ask, “What does the reader actually need from this?”

The Think, Brief, Write Model for Writing With AI

My preferred process is the Think, Brief, Write model.

It is deliberately simple because most AI writing problems happen when people skip the first two stages. They ask AI to write before they have worked out what they think, what the reader needs or what the message is trying to achieve.

That is why the output often sounds generic.

Stage 1: Think

Before asking AI to write anything, use it to improve your thinking.

Ask it questions. Challenge your assumptions. Explore alternatives. Upload notes, transcripts and rough ideas. Stress-test your argument.

This is where much of the value lives.

Useful prompts at this stage include:

• “What is the strongest argument in these notes?”
• “What feels unclear or unsupported?”
• “What assumptions am I making?”
• “What would a sceptical reader challenge?”
• “What is the one thing the reader needs to understand?”

This is not about polishing sentences. It is about sharpening the thinking behind them.

Stage 2: Brief

Once your thinking is clearer, brief AI properly.

When helping people think through AI-assisted writing, I often come back to four things:

For example, instead of saying:

“Write an email about our management training programme.”

Try:

“Write a kick-off email for our upcoming management training programme. The participants are emerging managers joining to support their personal and career development. Write it as an email. Keep the tone professional, positive and engaging.”

The difference isn’t subtle. The better the brief, the better the collaboration.

Stage 3: Write

Only now should you focus on creating the draft. And even then, your role isn’t finished.

You still need to apply judgement, remove unnecessary content, improve clarity, adjust the tone, add your own examples and opinions and make sure it still sounds like you.

That’s the human bit. And it is the bit that matters most.

In practice: use AI to think first, brief it clearly, then edit hard. The best AI-assisted writing should reduce the reader’s workload, not just the writer’s.

Use AI to Edit Your Writing

When using AI to edit your writing, one of the most useful shifts is to stop asking it to write immediately. Instead, ask it to critique.

For example:

• “Read this as a busy client. What is unclear or unnecessary?”
• “Tell me where this argument feels weak.”
• “What assumptions am I making?”
• “Where would a reader lose interest?”
• “Does this sound warm and direct, or too formal?”

One prompt I use a lot is:

“Before rewriting this, tell me what you think I am trying to say. Then tell me where the draft gets in the reader’s way.”

That second sentence matters. Often, the issue is not the writing itself. It is the friction the writing creates for the person reading it.

These prompts improve thinking before improving wording. And that’s often where the biggest gains come from.

This is also where Critical Thinking for AI matters. The better your judgement, the better your AI outputs become.

How to Tell When AI Has Taken Over

The biggest warning sign isn’t a particular phrase. It’s interchangeability.

If your writing could have been produced by anyone, it probably needs more of you in it.

Common signs include:

• sounding more formal than you normally would
• taking too long to get to the point
• using polished but empty language
• avoiding clear opinions
• adding unnecessary explanations
• feeling safe rather than useful

There are also certain words and phrases that often make me suspicious. Not because they are always wrong, but because they often appear when AI is doing too much of the driving.

Words like:

• unlock
• leverage
• delve
• seamless
• robust
• elevate
• transformative
• in today’s fast-paced world

The problem is not the individual word. The problem is when the writing starts to sound like it has put on a cheap suit and is trying to get through airport security.

Most people worry about AI getting things wrong. I’m increasingly convinced the bigger danger is that it gets things blandly right.

Writing in Your Own Voice Versus Your Company’s Voice

Sometimes you’re not writing purely as yourself. You’re writing on behalf of your organisation.

In those situations, AI can be helpful. Many AI tools now allow organisations to create custom instructions, assistants or skills that reflect company tone of voice, preferred language and brand guidelines.

Used well, these can help you draft content in line with company standards, check whether writing feels on-brand and identify language that doesn’t fit your organisation’s style.

For example, a simple company tone instruction might say:

“Use plain English. Be warm, clear and direct. Avoid corporate jargon. Keep sentences short. Do not over-polish the writing. We want to sound expert, but human.”

That kind of guidance gives AI something much more useful than “make this sound professional”.

However, the same principle still applies: AI can help, but people decide. A brand voice isn’t something to outsource entirely. It’s something to protect.

Read It Out Loud

This remains one of the simplest and most effective checks.

Once you’ve finished editing, read the piece out loud. Not silently. Properly out loud, even if you feel a bit ridiculous doing it.

You’ll immediately notice sentences that are too long, phrases you would never actually say, awkward wording, unnecessary formality and places where the writing has lost its personality.

The question isn’t, “Does this sound professional?”

The question is, “Does this sound like me?”

If the answer is no, keep editing.

The Human-in-the-Loop Writing Test

Before sending or publishing AI-assisted writing, ask:

• Is the main point clear?
• Have I removed anything the reader doesn’t need?
• Would I actually say these words out loud?
• Has AI made this more formal than it needs to be?
• Have I added my own examples or opinions?
• Does this reflect my judgement, not just information?
• Could this have been written by anyone?
• Have I done the final edit myself?

If not, the writing probably isn’t finished yet.

This is the human-in-the-loop test.

Not because the phrase sounds fancy. It doesn’t. It sounds like something an IT team might laminate. But the idea matters. AI can support the writing process, but a human still needs to make the final call on clarity, tone, relevance and usefulness.

The Real Skill Isn’t Prompting

There’s a lot of discussion about prompt engineering. Some of it is useful.

But I don’t think prompting is the real skill.

The real skill is judgement.

It’s knowing what matters, what can be removed, what the audience needs, when something sounds authentic and when a sentence is technically correct but completely forgettable.

AI can help you write. It can help you think. It can help you challenge assumptions and improve clarity.

But it can’t care on your behalf.

AI can’t know the relationship you have with a client. It can’t fully understand the dynamics of your team. It can’t decide what your organisation should sound like at its best. And it can’t tell you which parts of your voice are worth protecting.

That’s still your job.

So yes, use AI to write faster. Use it to challenge your thinking, sharpen your structure, test clarity and tone… but don’t use it to disappear from your own writing.

Because the goal isn’t to sound more like AI. It’s to sound more clearly like yourself.

Want Your Team to Write Better with AI?

Our Writing with AI workshop helps people use AI as a thinking partner, not a ghost writer. It focuses on better briefing, sharper judgement and clearer, more human writing.

You may also find our article on Beating Writer’s Block With AI useful if you want practical prompts to get started.

FAQs