Is Bad Writing the New Good Writing?

DEEP DIVE

I struggled to write this piece. Not because I lacked time or material or motivation, but because the writing tools at my disposal are now too good.

With GPT-5 and Opus 4.1 in the wild, every polished sentence risks sounding machine-made. Even a simple metaphor, when it lands too neatly, feels like a tell.

To address this, I’ve adjusted how I write. I’m now stricter with the fundamentals but messier with the finish (by which I mean I sabotage my own work in small, deliberate ways.)

When I mention this to people, they usually lean in. Which tells me 1.) I'm not alone in grappling with this strange new reality, and 2.) this topic deserves a newsletter.

Shall we 👇

We've crossed a bizarre threshold. AI prose has gone from “impressive for a computer” to better than most humans on their best day.

Crisp syntax, rich vocabulary, frictionless flow. This has become standard fare for the latest models.

The only weakness left is structural, what I’m calling the Connective Tissue Problem. AI can generate sentences that shine individually but often fail to connect into a unified whole.

Read any AI-generated essay and the cracks appear quickly: claims pile up without pause, transitions are scarce, and by the halfway mark the logic has unraveled.

That gap explains why a decent op-ed or research paper still requires hours upon hours of human effort. If we all had a “PhD in our pocket,” as Altman claims, wouldn’t we be seeing PhD-level papers everywhere?

Still, that structural gap isn’t what most readers feel day to day. Most encounters with AI writing come in short bursts – LinkedIn posts, newsletters, marketing copy – where connective tissue matters less and surface polish stands out more.

So forget the fear of “AI slop.” The likelier future is an internet saturated with relentlessly smooth, technically solid writing. Not always original or insightful (though there will be more of that too), but tidy, frictionless prose that often reads like today’s ChatGPT outputs.

When anyone with a ChatGPT account can produce near New Yorker-level smoothness, texture stands out. That’s why we’re seeing more and more people write poorly on purpose.

The lowercase rant on X:

The imperfect hook on LinkedIn:

The shot-from-the-hip newsletter edition:

The overly anecdotal SEO blog with the tragically flattened title:

These aren’t accidents. Rough edges carry a new kind of signal – proof that a human hand was actually in the mix.

I suspect this fad will pass. But for now, I’ve built a set of rules to help “humanize” my AI-assisted drafts.

Ironically, a lot of these involve doubling down on classic editorial principles, the kind you’ll find in books like this, this, and this. A handful, however, are deliberate hacks I use to break AI’s sheen.

Here they are (in an intentionally unkempt list) 👇

Mike and I agreed: the only honest way to honour the “bad writing” trend was to take it to its logical extreme. Which is why we’re so proud to announce BadWriterGPT, an AI that doesn’t polish your prose, it kinda ruins it.

For too long, AI has threatened to make writing unacceptably good and distressingly easy to read. BadWriterGPT boldly solves this problem by… lowering the bar, adding just enough flaws to ensure your copy sounds like it came from a fallible human.

Key features include:

The results are exactly as terrible – and oddly convincing – as you’d hope.

Try BadWriterGPT →

Right now, “bad writing” works as proof of humanity. I'd even argue it’s a reasonable response to AI’s sudden, unrelenting smoothness.

But it won’t last. Clarity has a gravitational pull that’s too strong.

Which is to say, most readers will continue to choose the shortest path to genuine insight irrespective of tech trends; stumbling through performative messiness isn’t that.

The deeper I've gone exploring this topic – experimenting with different ways to “humanize” my writing – the more I return to the same conclusion: the way to produce standout work in an AI world is to stop thinking about AI altogether.

Focus instead on what has always mattered – a sharp thesis, clean structure, and something worth saying. Good writing isn’t about the polish of the sentences or the tools behind them. It’s about the clarity and quality of your ideas.

So, my advice? Try the tactics outlined in this newsletter, test out BadWriterGPT, and have fun with this strange cultural moment. But if you’re building a serious editorial program, stick to the fundamentals – and join me in figuring out how to use AI to scale those instead.