When does your AI writing turn into slop?<br>Skip to main content

On this page

AI Version If youd rather read the AI version click here.<br>LLM writing is here, and that’s fine

There’s nothing inherently wrong with using AI in your writing process. I’d be a hypocrite if I spent all day burning through tokens as a software developer, then looked down on a blogger for using the same tools to write about a trip.

The appeal is obvious to anyone and im going to assume everyone reading this has used AI before. You give a machine your disorganized thoughts and get back something cleaner than what you gave it. Bloggers, lawyers, engineers, my dad, and government speech writers all use AI to write. It’s socially acceptable right now, and there’s so much of it online that you couldn’t avoid it if you tried.

I read dozens of Hacker News posts a day and a noticable amount of them feel like rough notes dropped into a frontier model and published straight to a tech blog. I’m not the only person who has noticed this.

Ai-isms

But what do we mean exactly by noticable? People have started calling these cliche AI writing patterns “AI-isms”. Im sure everyone is familiar with the Its not X its Y” pattern that started to proliferate everywhere in 2025. What do you feel when you see it? I personally feel my trust in what im reading diminish a bit. I usually at that point either skip it, or I start to wonder about what was fed into the LLM to produce this output. Involuntarily im taken out of what I was reading, and I start to consider what the user prompted into the LLM to produce this output. Sometimes this is unfortunately a required step because the writer gave it a few too many “Make it more concise” loops haha.

The Turing Window<br>The Turing Window is the brief stretch of time certain AI writing goes unnoticed by human readers. Writing outside this window is a negative quality indicator.<br>Honestly, the name is a stretch.

These sentences and phrases on their own may convey your message or idea but when the reader has seen that exact phrasing or framing 10 times that morning, it lessens the impact of your writing. Millions of people interact with chatbots constantly, they see AI-generated emails, blog posts, LinkedIn updates, documentation, marketing copy, and social media posts. They are aware of the trends of writing even as passive consumers of it, and it truly does affect their perception of what they are reading. A phrase stops sounding insightful and instead presents itself as a lazy shortcut. The moment a reader starts noticing the writing process instead of the idea being communicated, the author has already lost some credibility.

That’s why AI writing ages differently than human writing. A human cliché usually dates a piece to a genre, profession, or decade. An AI-ism dates it to a model generation.

What are the modern AI-isms then?

Its mid 2026 and LLM writing models are getting better and better. But they are still not perfect. A while ago I read this post from OpenClaw:

So we started moving things out of core: channels, providers, heavy tools, parsers, optional integrations. The plugin inventory shows what still ships in core, what installs separately, and what is source-checkout only.

The problem: I underestimated how difficult it would be to get this right. For a few releases we ended up in the worst middle state: too much moved toward plugins, while too many plugins were still bundled, repaired, staged, checked, or dependency-loaded in places users feel immediately.

These two paragraphs seem to have every single AI-ism that is currently popular. The writing is compressed until every noun has to do the work of a sentence, extreme overuse of colons, sentences that are short and punchy without depth, and verbs carry implications that are never expanded on. There are also strange compounds such as “source-checkout only.” The detail is packed so tightly that it’s hard to tell what actually happened. It reads like the result of a dozen requests to “make this more concise.”

I’d say at this point we are still in the window for this style of writing unless you are a major consumer of tech blogs and are nitpicker like me.

For an example of something rapidly exiting the window, there is this recent blog post from a random biking company I came across about a bike-fitting tool:

That’s the part worth pausing on. No wind tunnel booking, no six-figure lab, no all-day session — just a rider, a fitter, a bike, and a platform that turned a handful of photos and a calibrated setup into a working aero model.

This type of writing is just a longer way of saying “Its not X, its Y”.

Whats the fix?

Despite my negative tone here, im not telling people to stop using AI to write. In fact, I have exactly 0 problem with people doing the exact same notes -> LLM -> polished writing process. What I suggest is that everyone take it just one step further and EDIT THE AI-ISM OUT . You can keep all the exact same content, but taking 10...