Let's write a blog using my tone:

• The topic is how to use LLMs as a writing tool
• My opinion is: it is a great tool for articulating thoughts
• Adding facts: let's research how AI and LLMs now most of the internet content
• Adding facts: let's research what makes a copy personal and engaging
• Let's highlight areas where we can add humor
• The audience is on Medium and mostly cares about self-improvement and career advancement in marketing
• I have an MD file generated before that has my tone and other voice related things. 

Here is the output: 

 

Everyone's Using AI to Write. Almost No One's Thinking While They Do.

Scroll LinkedIn for five minutes. Count how many posts open with "In today's fast-paced world" or end with three bolded takeaways nobody asked for.

That's not writers losing their voice. It's AI doing exactly what people ask it to do: produce content before anyone has had a thought worth expressing.

The numbers back up what your feed already tells you. Ahrefs looked at 900,000 newly created web pages in April 2025 and found AI-generated text in 74.2% of them, mostly a human-AI mix, not pure AI output. Europol estimated in 2023 that up to 90% of online content could be synthetically generated by 2026. And nearly 94% of marketers say they plan to use AI for content creation, up from roughly a third two years ago.

Almost nobody is opting out.

My problem isn't that people use AI to write. It's that most of them use it to generate before they've figured out what they want to say.

Feed a model a topic and it hands you paragraphs. No opinion went in, so no opinion comes out. Readers can feel that within a sentence or two.

Ask an LLM for its take on your pricing strategy and it will happily agree with you. Open another tab, argue the opposite case, and it will agree with that too. It isn't lying. It just doesn't have a position until you give it one.

What makes writing land

Writing lands when it has a point of view.

It names a competitor's real strength instead of pretending they have none. It explains the reasoning behind a decision instead of just announcing it. It admits the trade-off instead of hiding it.

A model can imitate the shape of those sentences. It can't supply the conviction behind them. It didn't make the decision. It doesn't have to live with the consequences. You do.

That's what people get backwards. They open ChatGPT with a blank prompt and ask it to write the post.

Of course it comes out generic. Nothing specific went in.

The way I actually use it

I don't ask it to write. I talk to it.

I dump the raw version of what I think: half-formed, out of order, probably with a typo or two. Then I ask it to organize it back to me.

I argue with it. I ask where the logic is thin, where I'm repeating myself, where I'm hiding behind a vague word instead of saying the specific thing.

What comes back is still my thinking. It's just less tangled.

That's where an LLM earns its keep: as a fast, patient editor for a thought you already have. Not a replacement for having one.

Why this matters if you work in marketing

Everyone in your feed is publishing some version of the same AI-smoothed post.

The floor for technically competent writing has dropped to almost nothing. The ceiling hasn't moved, because it was never about sentence quality. It was about having a real opinion and being willing to put it in writing.

The person who shows up with a point of view, sharpened by AI, not replaced by it, reads like the only human in a room full of very polite chatbots.

That's a content advantage. It's also a career advantage.

Nobody remembers the account that posted something. People remember the person who said something. I've seen how differently friends engage with posts that actually manage to convey real thought and feeling.

Use the tool to think out loud, write, rewrite, edit, and move ahead. Not to think for you.

LLMs like ChatGPT and Claude are tools to help people think and write better, and in this specific area, they're the new pen and typewriter.

Disclaimer: an LLM helped shape this piece, which is exactly the point. I brought the opinion and argued it out; AI helped organize and articulate it.