4,000 followers. Gone. Deleted the account completely.

And as a ghostwriter who spends my days fighting against generic AI content, I need to tell you why this matters.

Because what he exposed isn’t just about LinkedIn.

It’s about what happens when platforms optimise for average.

Microsoft owns LinkedIn.

Microsoft invested $13 billion in OpenAI.

OpenAI invested in Nvidia.

Now LinkedIn offers you a button: “Write with AI.”

You think it’s there to help you write faster.

It’s not.

Every time you click it: → Your post becomes training data for OpenAI → OpenAI’s model improves → Microsoft’s investment grows → Nvidia sells more chips

You’re not using a tool. You’re fueling an investment loop.

And the output? Perfectly average. By design.

Because that’s what LLMs do. They aggregate patterns. They find the statistical middle. They give you what’s already been said a thousand times.

That’s the opposite of what good ghostwriting does.

My job is voice accuracy.

Making content that sounds like my client wrote it. Not like I wrote it. And definitely not like a robot wrote it.

That’s the work.

I’m not a content generator. I’m a voice replicator with taste filters.

And I can tell you from hundreds of client posts: LinkedIn’s AI button is working against that goal.

Every time a founder clicks it, their voice gets a little more generic. A little more corporate. A little more forgettable.

Then they wonder why engagement drops.

Tom called LinkedIn’s AI feature a Ponzi scheme.

Strong words. But he’s not wrong.

Platforms don’t optimise for your success. They optimise for their business model.

LinkedIn’s business model now includes:

Your unique voice? Your authentic perspective? Your ability to stand out?

That’s not in their KPIs.

So when they give you an AI button that makes you sound like everyone else, that’s a feature for them. Not for you.

The bad ghostwriters are using LinkedIn’s AI button. Pumping out generic content for clients. Fast. Cheap. Forgettable.

They’ll lose those clients in 6 months when engagement tanks.

The good ghostwriters are doing voice profiling. Building banned word lists. Creating signature phrase libraries. Editing until the AI fingerprint is gone.

They keep clients for years.

I’ve built entire frameworks around this because I had to.

When every founder you work with expects you to “just use ChatGPT,” you need systems that prove voice accuracy beats speed.

I don’t use LinkedIn’s AI button. Ever.

Here’s my actual workflow:

I extract:

This becomes the voice profile. The blueprint.

Without this, I’m just guessing. And guessing sounds generic.

Tools like Wispr Flow. I voice-draft based on their profile, or capture their actual speaking.

Why? Because you can’t fake authenticity when someone speaks naturally.

The patterns, the pauses, the specific words they choose when not typing. That’s where voice lives.

ChatGPT is declining for creative writing. I’ve tested this extensively.

Claude Opus 4.5 maintains voice better while cleaning structure. It doesn’t default to corporate speak as aggressively.

But here’s the key: I feed it the voice profile first. Every time.

Context beats prompts. Always.

This is where most ghostwriters fail. They think AI output is “done.”

It’s not.

I run every piece through:

If that last one is yes, I start over.

Before delivery, every piece gets:

If any box is unchecked → revise.

This takes time. But it’s why clients stay.

AI gives you templates.

Ghostwriting gives you taste.

Templates are: “[Number] ways to [achieve outcome] without [common objection]”

Taste is: Understanding that this client never uses numbered lists. They tell stories. They’re confrontational but never mean. They include screenshots, not stock photos. They end with questions, not CTAs.

Taste is the ability to say no to everything and yes to a few things that define the right path.

And here’s the thing about taste: AI doesn’t have it.

AI has averages. Patterns. Statistical middles.

Taste is human. Taste is what makes content worth reading.

As a ghostwriter, my taste filters are trained on one person: the client.

Not on “best practices.” Not on “what performs well.” On what sounds like them.

They’re optimising for speed.

Wrong question.

The right question:

Because here’s what happens with the speed approach:

Month 1: Client is thrilled (wow, so fast!)

Month 2: Client notices it doesn’t quite sound like them

Month 3: Engagement drops

Month 4: Client leaves

I’ve seen this dozens of times.

The ghostwriters who last are the ones who build systems for voice accuracy, not content velocity.

When I profile a client’s voice, I’m looking for:

Sentence-level patterns:

Vocabulary patterns:

Structural patterns:

Emotional patterns:

Belief patterns:

This isn’t a template. This is a fingerprint.

And once you have it, you can replicate voice at scale without losing accuracy.

Ghostwriting isn’t about writing.

It’s about disappearing.

The best ghostwriting is invisible. The reader thinks the client wrote it. They should never think, “I bet they hired someone.”

Deep listening. I spend more time in client kickoff calls than actually writing. I’m listening for speech patterns. Vocabulary. Energy. What makes them angry. What makes them laugh.

Obsessive research. I read everything they’ve written. Watch their videos. Listen to their podcasts. Track their recent LinkedIn activity. What’s happening in their life right now that should show up in their voice?

Psychological profiling. What’s their relationship with their audience? Teacher? Equal? Challenger? Different tones for different relationships.

Quality control paranoia. Before anything ships, I ask: “If I removed their name, would I know who wrote this?” If no, I failed.

This work is invisible to clients. They see the output. They don’t see the 10 revision passes that made it sound effortless.

That’s the job.

They think they’re hiring a writer.

They’re not.

They’re hiring a voice replicator with strategic filters.

The value isn’t in “I can write LinkedIn posts.”

The value is in “I can write 30 LinkedIn posts that all sound exactly like you, even though you didn’t write them, and your audience will never know the difference.”

That’s a completely different skill.

And it’s incompatible with clicking AI buttons that optimise for the average.

Stop.

Just stop.

Every click:

Tom Vogel figured this out after 20 years and 4,000 followers.

You don’t have to delete LinkedIn.

But you do have to stop letting it write for you.

Build your own voice profile:

Then use AI with that profile as context.

Don’t ask it to “write a LinkedIn post.” Ask it to “write in this specific voice using these patterns.”

Context beats prompts. Always.

Don’t hire based on speed.

Hire based on their voice profiling process.

Ask them:

If they say “I use ChatGPT,” run.

Stop using client-agnostic templates.

Build voice extraction frameworks. Build systematic editing processes. Build quality control checklists.

The ghostwriters who survive the AI wave won’t be the fastest.

They’ll be the ones who can replicate voice at 85%+ accuracy across unlimited content.

That’s the skill AI can’t replace.

Are you optimising for the platform?

Or are you optimising for your audience?

Because you can’t do both.

LinkedIn wants you to be average. Generic. Scrollable.

Your audience wants you to be specific. Different. Memorable.

As a ghostwriter, I’m not here to make my clients sound good.

I’m here to make them sound like themselves.

And that’s the opposite of clicking a button that makes everyone sound the same.

Your move.

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

Find me here → linkedin.com/in/meetsarra