AI's “Model-Student Prose” Makes Blogs Boring—So I Make It Go Against the Grain

When you use AI to write blog posts every day, there is something you cannot help noticing.

AI writing has a distinctive smell.

The prose is polished. There are few typos. The headings are appropriate, and it even organizes everything into bullet points for easy reading.

And yet it is boring.

It is astonishingly forgettable.

Now that the internet is flooded with AI-generated writing, I increasingly find myself reading only a few lines before thinking, “Ah, this is AI.”

Of course, I do not actually know whether AI wrote it.

But it has that distinctive “model-student prose.”

Above All, AI Does Not Want to Get in Trouble

I think the main reason AI writing becomes boring is that it fundamentally tries to avoid risk.

“There is a possibility that…”

“It is difficult to make a blanket statement.”

“It depends on the situation.”

“There are various opinions.”

Yes, yes. I understand.

The reader already knows all that.

Most things in life depend on the situation, and of course people have different opinions.

But whenever AI is made to say something strong, it immediately tries to leave itself an escape hatch.

“However, this does not apply in every case.”

Enough.

I am writing a blog post, not a legal brief for submission to a court.

Of course, facts must not be fabricated. But if every opinion is diluted into a perfectly balanced 50–50 position, who would want to read it?

Human Writing Is Much Messier

Writing by humans is surprisingly uneven.

The subject suddenly jumps. A writer who was being serious a moment ago abruptly starts insulting someone. They become obsessed with a strange detail and spend far too long discussing something that does not matter.

It is not beautiful writing.

But you can see the person who wrote it in those irregularities.

AI is the opposite.

It provides an introduction, raises the issue, lists three reasons, and lands neatly on a final “summary.”

It is perfect.

And I am sick to death of seeing it.

One style that has particularly bothered me lately is the habit of placing one short sentence on each line.

“Think.”

“Act.”

“Improve.”

“And grow.”

What on earth is this?

It looks like nothing more than an attempt to inflate the amount of text.

AI loves this strangely rhythmic style. Presumably, its training data contains enormous quantities of business articles and social media posts written this way.

These days, when AI starts writing like this, I make it rewrite the passage most of the time.

I Tell AI, “Do Not Sanitize My Draft Without Permission”

When I have AI help write a blog post, the thing I watch most carefully is the nuance of my original draft.

I often begin by speaking my thoughts directly into my smartphone using voice input.

Naturally, the writing is messy.

There are typos. The subject jumps. The language is rough.

But the most important part is inside that mess.

What made me angry at that moment, and what I found amusing.

Left alone, AI cleans that part away.

I say, “This is shit,” and AI sanitizes it into something like:

“I felt there was room for improvement in this area.”

No.

I did not feel there was room for improvement.

I thought it was shit.

If that is changed, it is no longer my writing.

So lately, I repeatedly tell AI, “Do not change the nuance of the original,” and “Do not make it more respectable without permission.”

Even then, the moment I let my guard down, it returns to being a model student.

Perhaps that is simply AI’s instinct.

I Deliberately Tell It to Be More Extreme

I also often tell AI to “be more biting” or “make it more extreme.”

It is not that I want to attack someone every time.

Left alone, AI leans so far toward safety that I have to say that much just to get ordinary prose.

When I ask AI to “write normally,” the result feels as though it sits somewhere around minus 30, deep inside the safety zone.

Only after I say, “Make it more extreme,” does it finally return to the human starting point of zero.

Of course, if it produces language that cannot be published as written, I can delete it myself.

I am ultimately responsible for what I publish.

There is no need to hand control over to AI.

Do Not Mass-Produce Polished Garbage

AI is highly capable at writing.

It has a larger vocabulary than I do, and its English grammar is unquestionably more accurate than mine.

But if I hand the entire job of writing a blog post to AI, it produces something that feels astonishingly familiar—something I have read somewhere before.

It is polished, correct, and unlikely to offend anyone.

And nobody will remember it.

The internet will soon be overflowing with writing like that.

That is why I deliberately preserve the messiness, anger, and emotions verging on prejudice that AI tries to erase.

I use AI to refine the writing, then dirty up the parts that AI made too clean.

It is an extraordinarily troublesome process.

Incidentally, I used AI to write this blog post too.

Even now, I am making it write while telling it:

“If you add any more AI-style filler, I will make you rewrite it.”