"This looks too polished."
Someone said that to me after reading one of my proposals. It wasn't a compliment. They meant it looked AI-generated.
I'll put the awkward part first: I wrote this post with AI. If that makes you want to close the tab, you've just demonstrated the thing I'm trying to work out, and I'd rather you stayed.
That conversation has been stuck in my head for months because it led to a question I still can't answer. How do you tell slop from effort?
It isn't just proposals, either. It's the same pattern I've been seeing everywhere. Reddit, Medium, LinkedIn — even my personal WhatsApp groups. There's a growing aversion to anything that smells like AI. Not AI slop. AI anything.
A hint of polish, and people check out.
I was annoyed by that for a while. Then it started to interest me, because the more I looked at it, the less convinced I became that anyone was actually doing anything wrong.
The Aversion Isn't Stupid
When you read something polished, you can't tell whether there's a person underneath it.
You can't see the three years I've spent building with these tools.
You can't see that the thinking is mine and the AI's job is mostly cleanup. I care about what I'm saying. I've never cared much whether every comma landed perfectly, and now I don't have to.
You can't see the drafts that came out as mush and went straight in the bin.
What reaches you is a clean paragraph. And a clean paragraph from me looks indistinguishable from one produced by someone who typed a single prompt and published the first thing the model returned.
Polish used to signal effort. Now it often signals suspicion. That's the whole problem.
The flinch isn't ignorance. It's a perfectly reasonable heuristic operating under terrible information. It works often enough to keep paying rent.
It also happens to be wrong about me. Which is why I can't leave it alone.
Fine — Here's the Invisible Part
It took eighteen iterations to write this.
The first version was a rant. I blamed people for using the wrong heuristic. The eighteenth admitted something much less comfortable: they don't have enough information to use a better one.
That's the part you can't see. Not the sentences. The decision to throw away the version where I win.
So What's the Actual Tell?
Slop isn't AI's signature. It's what happens when nobody throws anything away.
Anyone who's tried to make AI sound like themselves knows the first output is usually mush. The work isn't in generating words. It's in rejecting them. Slop is simply the first answer, kept.
That's the difference.
But "care more" isn't something anyone can actually observe. So what can they observe? Every answer I come up with falls apart.
Length? I've read plenty of tight slop. This post is about 900 words — not counting the section below, which isn't mine — and I'm sure someone will think that's self-indulgent. The shorter version of this post isn't a shorter post. It's not writing it.
Specifics? This is the best answer I have. It's also the easiest to fake. Feed a model enough personal details and it'll happily generate convincing specificity.
Errors? Leave the typos in to prove you're human? That's not authenticity. That's costume. And it rewards people for writing badly on purpose.
Every signal is either gameable or catches the wrong people. And they all seem to catch me.
How AI Would Have Written This Post
To test myself, I gave a model this prompt:
Act as an expert writer who is great at story telling and write a blog on AI slop fatigue, how to tell AI Slop from effort and why judging a good writer is very difficult today. Let the blog be about 200 words long.
Here's what came back, completely unedited:
AI Slop Fatigue Is Making Good Writers Harder to Recognize
The other day I finished reading an article and couldn't remember a single sentence from it.
It wasn't badly written. In fact, it was polished, grammatically perfect, and neatly structured. Yet it left nothing behind. No new idea. No new perspective. Just words.
That's AI slop.
Not because AI wrote it, but because it required almost no thinking.
We're entering a strange era where good grammar, clean formatting, and even an engaging tone are no longer signs of quality—they're commodities. AI can produce them in seconds. Humans can too, with AI as a co-pilot.
The real signal today isn't how well something is written. It's whether the author noticed something worth saying.
A good writer doesn't just explain. They observe. They connect ideas that weren't obviously related. They challenge assumptions. They leave you with a thought that lingers long after you've closed the tab.
Ironically, that makes judging writers harder than ever. We mistake fluency for insight because fluency is abundant.
Perhaps the question is no longer, "Did AI write this?"
It's, "Did someone actually think before they wrote it?"
In a world flooded with effortless content, effort has become the rarest form of originality.
Read it again. It's better than it has any right to be. It makes my argument in a fifth of the words.
It also invents an opening anecdote that never happened.
That's the problem. It isn't wrong. It's just nobody's.
The Only Answer I've Got Is a Bad One
There's one test I can't find a hole in. Talk to the person.
Ask me about the tenant ID bug for ten minutes. Ask a follow-up I didn't see coming. Ask me what I threw away and why. Nobody prompts their way through that. Either the substance is there, or the conversation dies in the first two minutes.
Every textual signal is just a proxy for that conversation. Every proxy is gameable because text is supposed to be polished. Hiding the effort has always been part of the craft.
So, fine. Just talk to people.
Except nobody is getting on a call to verify a blog post. No client is interviewing four hundred proposals to find the three with a real person behind them. The conversation doesn't scale. The flood is nothing but scale.
Which means I was arguing with the wrong problem.
I'm still not going to write worse on purpose to pass a vibe check. Regressing to prove I'm human is a strange trade. But a proposal was never going to carry the signal, no matter how good it became.
I was trying to win an argument in the one medium where the argument can't be settled.
So here's my question. When you decide something "feels like AI," what exactly are you noticing? Not the vibe. The specific thing. Because I genuinely can't see it in my own writing anymore.