It can’t. It won’t. And the 200% surge in executives using AI ghostwriting tools (cheers, TechCrunch) is creating an avalanche of content that sounds like a LinkedIn bot had a breakdown at a TED Talk.

I need to be honest with you about something uncomfortable.

I use AI. Every single day. Not to write my clients’ content, but to excavate it. There’s a difference, and that difference is worth approximately £4,997 in positioning that most ghostwriters are pissing away because they don’t understand what AI actually does.

Let me show you where everyone’s getting this catastrophically wrong.

Last month, Marcus hired a ghostwriter. Paid £1,200 for a month of LinkedIn content. Three posts a week. Professionally formatted. Engagement-optimised. Completely fucking soulless.

Every post sounded like it was written by compliance, edited by legal, and approved by someone who thinks “synergy” is a personality trait.

Marcus read them and felt nothing.

His audience read them and scrolled past.

The ghostwriter? Delighted with himself: “I used AI to optimise for the algorithm!”

The algorithm doesn’t buy from you. Humans do.

Here’s what that ghostwriter missed: AI isn’t a replacement for thinking. It’s a mirror.

Whatever mediocre prompt you give, it reflects back mediocre content. If you ask it to “write a LinkedIn post about leadership,” you get the same generic slop that 10,000 other people are posting this week.

But if you give it your actual thinking, the rant you went on in a client call, the contrarian take you’re too scared to publish, the framework you’ve been using for years but never named, THEN it becomes useful. Not as a writer. As a translator.

I’ll tell you exactly how I use AI in my ghostwriting process. Not because I’m trying to be transparent. Because watching people use it wrong makes me want to fight someone.

Phase 1: Voice Archaeology

I record a 60-90 minute conversation with my client. We’re not talking about “content strategy.” We’re talking about the last time they wanted to quit. The client who made them furious. The moment they realised their entire industry was doing it wrong.

I’m excavating. Finding the 47-second moment where they dropped their corporate mask and said what they actually think.

Then, and only then, I feed that transcript to AI with a single instruction, something like this (I’m not giving away the playbook): “Extract the contrarian beliefs, the specific examples, and the visceral language. Give me the raw material.”

AI finds the gold I might have missed. The throwaway line that’s actually the hook. The metaphor they used, without realising it, was brilliant. The rage underneath the professional politeness.

But it doesn’t write the post. It gives me the ingredients.

Phase 2: Translation, Not Generation

Now I have their actual voice, not the LinkedIn-bot version they’ve been trained to write in for 15 years. I structure it. I punch it up. I make sure the rhythm matches how they actually talk.

Then I show AI the structure and ask: “Find me three different ways to open this that stop the scroll.”

It gives me options:

I pick option 3. Because I already knew that was the right hook, I just needed AI to show me I wasn’t crazy.

See the difference? I’m not asking AI to write. I’m asking it to pressure test my instincts.

To give me variations.

To find patterns in my client’s speech that their corporate conditioning has trained them to hide.

Phase 3: The QA Filter AI Can’t Do

Here’s where everyone fucks it up.

They get their AI-generated content, read it once, think “this sounds pretty good,” and hit publish.

No. Stop. That content sounds like 10,000 other posts because you didn’t do the actual work: making it sound like your client specifically.

I run every draft through what I call the “Anti Generic Test”:

Could any of my client’s competitors have written this exact post?

If yes, it’s beige. Doesn’t matter how well-structured it is. Doesn’t matter if it got AI approval. If I remove their name and it could belong to anyone, I’ve failed.

This is where human judgment lives. AI can’t tell you if something is distinctly yours. It can only tell you if something is grammatically correct and topically relevant.

That’s not good enough.

Let’s talk about the thing everyone’s pretending isn’t happening.

Some ghostwriters are using AI to churn out content, slapping their client’s name on it, and charging £2,500/month. The client thinks they hired a human. They hired an algorithm with a LinkedIn profile.

That’s not ghostwriting. That’s fraud with extra steps.

When Wired caught an AI tool plagiarising Forbes verbatim, everyone clutched their pearls. “How could this happen?!”

Because you’re treating AI like magic instead of machinery. It doesn’t “understand” plagiarism. It pattern matches. If you don’t know how to pressure test its output against originality, you’re publishing stolen goods.

I tell every client exactly how I use AI. Not because I’m noble. Because if they ever discover I used a tool I didn’t disclose, I’ve lost their trust forever. And trust is the only thing I’m actually selling.

Here’s my rule: If AI touched it, I tell you. If I used it to transcribe, to extract themes, to generate options, you know. Because you’re not hiring me to hide tools from you. You’re hiring me to use tools better than you can.

There are exactly four uses where AI doesn’t make you sound like a bot:

1. Transcription and theme extraction. Record yourself talking about your work for an hour. Feed it to AI. Ask it to find recurring themes, contrarian takes, and specific examples. This is brutally effective because you’re not asking it to generate, you’re asking it to organise what you already said.

2. Variation generation. You’ve written a hook. You know it’s good. Ask AI for 10 variations. Not because you’ll use them. Seeing alternatives helps you confirm your instinct or sparks a better direction.

3. Structural scaffolding: Give AI your messy brain dump and ask it to organise into: Hook → Story → Insight → CTA. Then rewrite every single sentence in your actual voice. The structure helps. The sentences? All you.

4. Editing, not writing. Your draft is done. Feed it to AI and ask: “Where is this generic? Where does this sound corporate? What would make this more specific?” Use its feedback to sharpen. Ignore its suggestions to make things “more professional” (that’s code for more beige).

Notice what’s missing from that list? “Write my content for me.”

Because AI can’t do that. Not well.

Not in a way that builds authority.

Not in a way that makes people think “I need to work with this person specifically.”

Speed to narrative beats perfection paralysis.

Everyone’s obsessed with whether AI content is “as good as” human content.

Wrong question.

The right question: Does this content sound unmistakably like YOU, move someone closer to buying, and get published this week instead of sitting in drafts for three months?

If you’re using AI to overcome perfectionism and actually ship, good. If you’re using AI because you think it’ll make you sound smarter, you’re making yourself forgettable.

I’ve closed many ghostwriting contracts this year. Every single client came from content I wrote using AI as a tool, not a replacement.

They didn’t hire me because my posts were perfectly polished.

They hired me because my posts sounded like a human who had opinions, edges, and a distinct POV.

That’s not something AI generates. That’s something you excavate, amplify, and refuse to sand down for mass appeal.

Thanks for reading SubText! This post is public so feel free to share it.

If you’re burning 10 hours a week writing content that gets 47 likes and zero pipeline, you have three options:

Option 1: Keep doing it yourself. Manually. Slowly. Burning revenue-generating hours on tasks you hate. (This is martyrdom, not strategy.)

Option 2: Hire a ghostwriter who uses AI as a shortcut to churn out beige content faster. You’ll get consistency. You’ll get posts. You won’t get a pipeline. (This is outsourcing mediocrity.)

Option 3: Work with someone who uses AI as a voice excavation tool, not a replacement for thinking. Someone who records your actual thoughts, extracts the contrarian beliefs, structures them into scroll-stopping content, and makes sure every post sounds unmistakably like you. (This is the thing I do.)

On one side: Founders who understand AI is a tool for amplification, not substitution. Who uses it to move faster without sounding generic. Who protect their voice while killing it with technology.

On the other: People who think they can prompt-engineer their way to authority. Those who believe “write like a human” is sufficient instruction. Who doesn’t realise their content sounds exactly like everyone else’s because they’re all using the same mediocre prompts.

I know which side closes deals.

Still me. Still using AI. Still making sure you can’t tell.

If you want to see exactly where your LinkedIn content is bleeding authenticity (and pipeline), take my 10-minute Founder Voice Audit. It’s free and I’ll give you a steer.

DM me “GHOST” if you’re ready to fix it.

Fair warning: I only work with people who understand the difference between delegation and abdication.

If you want someone to “just handle content,” there’s an AI tool for that.

If you want someone to make you sound braver than you currently are? Talk to me.

SubText Written by Sarra, the Ghost, professional ghostwriter, strategist, and marketer hiding in plain sight behind the voices you read every day.

Find me here → linkedin.com/in/meetsarra