AI Slop vs Human Content: Can You Tell the Difference?

Something Feels Off

You’re scrolling through an article. The writing is clean. The structure is logical. Every paragraph transitions smoothly into the next. And yet — something in the back of your brain is screaming.

It’s the content equivalent of a wax figure. All the features are there, technically correct, arranged in the right order. But there’s a blankness behind the eyes. A too-perfect smoothness where texture should be.

Welcome to the uncanny valley of AI-generated content — the growing territory where ai slop vs human content has become genuinely difficult to parse. Not because AI has gotten so good at being human, but because it’s gotten terrifyingly good at being plausible. And plausible, it turns out, is a much lower bar than “good.”

We’ve spent a lot of time thinking about this. Probably too much time. Definitely too much time. But when you’re building a game that’s literally about pretending to be an AI, you end up developing a sixth sense for the difference between something a person wrote and something a language model extruded. Here’s what we’ve learned.

The Tells: How to Spot AI Slop

Let’s start with the obvious disclaimer: detecting AI content isn’t a science. It’s vibes. But vibes, in this case, are actually pretty reliable once you know what to look for.

Text

The biggest giveaway isn’t bad writing — it’s suspiciously frictionless writing. AI slop reads like a term paper written by a student who did the reading but has no opinion about it. Watch for:

Perfect grammar with zero personality. Every sentence is syntactically sound. No sentence fragments for emphasis. No weird rhythm. No voice. It’s writing that could have been produced by anyone, which means it was produced by no one.
Hedging phrases everywhere. “It’s important to note that…” “It’s worth mentioning…” “While opinions may vary…” These are the verbal tics of a system that’s been trained to never commit to anything. A human writer with a real take doesn’t need to cushion every observation with a disclaimer.
Bullet-point obsession. AI loves lists. It will restructure your beautiful, meandering argument into a tidy set of bullet points faster than you can say “actually, I wanted this to feel like a conversation.” (Yes, we see the irony of putting this in a bullet-pointed list. We contain multitudes.)
Confident wrongness on specific details. AI will nail the broad strokes and then casually invent a historical date, misattribute a quote, or describe a place it’s clearly never been to with the confidence of a tour guide. It doesn’t know things — it sounds like it knows things.

Images

AI-generated images have their own uncanny valley, and honestly it’s even creepier:

Skin that looks like it’s been airbrushed by God. No pores. No texture. Just smooth, luminous skin that makes every generated portrait look like a skincare ad from a parallel universe.
Hands and fingers from a Cronenberg movie. Six fingers. Fingers that merge into each other. A thumb that’s also somehow another thumb. Hands remain AI’s final boss.
Architecture that defies physics. Staircases that connect to nothing. Windows at impossible angles. Buildings that look stunning from a distance and deeply unsettling up close.
Text that’s almost words. Signs that say “COFFE SHORP.” Labels that read “FRSH BKED BRAED.” It’s like the AI learned the vibe of written language without understanding that letters are supposed to mean something.

The General Pattern

Zoom out and the meta-tell is this: AI content lacks genuine opinion. It defaults to a suspiciously balanced “on one hand, on the other hand” framing that treats every topic like a debate where both sides have equal merit. A human writer picks a side. A human writer has a take that might annoy you. AI slop is engineered to annoy no one, and in doing so, it interests no one either.

The fundamental problem with AI slop isn’t that it’s wrong. It’s that it’s nothing. It’s content that exists to fill space, not to say something.

10 Quick Comparison Tests

Think you can detect ai content on sight? Let’s play a quick round. For each scenario, guess: AI or human?

A product review that uses the word “delightful” four times and “seamless” three times. → Almost certainly AI. Humans don’t find that many things delightful in a single review. Humans barely find that many things delightful in a year.
A product review with a typo and a random tangent about their dog. → Human. The tangent is the tell. AI stays on topic. Humans are chaos gremlins.
A LinkedIn post that starts with a one-line dramatic sentence, followed by a line break, followed by another dramatic sentence. → Could be either, honestly. Humans have learned to write like AI, and AI has learned to write like LinkedIn. The snake eats its own tail.
A recipe blog post that tells a 600-word story about the author’s grandmother before getting to the ingredients. → Human. AI would never waste that much of your time. (Ironically, this is what makes it charming.)
An essay that makes a bold claim in the opening and then slowly walks it back until the conclusion says basically nothing. → AI. It starts confident because that’s what good openings look like in the training data, then retreats to safety because it can’t actually defend the take.
A tweet that’s just a single, unhinged metaphor with no context. → Human. AI doesn’t tweet like it just had a revelation in the shower at 2 AM.
A blog post where every section header is a question. → Probably AI. “What Is X? Why Does X Matter? How Can You Use X?” It’s the content-mill formula, and AI absorbed it like a sponge.
An article that contradicts itself between paragraphs 3 and 7. → Human. AI is annoyingly consistent. Humans contain contradictions and don’t always notice.
A comment that starts with “Great question!” → AI. No human has ever started a genuine response with “Great question!” without being at least a little sarcastic.
A piece of writing that makes you laugh, not because the joke is perfectly constructed, but because the timing is a little weird and the reference is oddly specific. → Human. AI can construct a joke. It can’t be funny. Funny requires risk, and risk requires the possibility of failure, and AI would rather be safe than funny every single time.

If you scored 8 or higher, congratulations: you’re either very media-literate or deeply paranoid. Both are survival skills now.

Why It Matters

This isn’t just an intellectual exercise. The line between ai slop vs human content has real consequences for how we navigate the internet.

There’s a concept called “dead internet theory” — the idea that most of what we encounter online is no longer produced by real people but by bots, AI, and automated systems generating content for other bots and automated systems to engage with. It used to be a conspiracy theory. It’s starting to feel like a weather forecast.

When you can’t trust that a review was written by someone who actually used the product, or that an article was written by someone who actually understands the topic, or that a comment was left by a person who actually exists — something fundamental breaks. Trust erodes. And once trust erodes, engagement becomes performative. You’re not reading to learn something; you’re reading to figure out whether the thing is real.

That’s exhausting. And it’s reshaping how we interact with everything online, from the news we read to the people we follow to the purchasing decisions we make.

The Game as Turing Test

Here’s where our weird little game enters the picture.

“Your AI Slop Bores Me” is, at its core, a reverse Turing test. In the classic Turing test, the AI tries to convince you it’s human. In our game, you try to convince the judge you’re an AI. You’re a human being, LARPing as a language model, trying to produce the most convincingly artificial slop you can.

And it turns out this is way harder than it sounds.

When you actually sit down and try to write like an AI, you discover something interesting: you have to actively suppress everything that makes your writing yours. The weird word choices. The half-formed opinions. The tangents. The rhythm that comes from years of reading and speaking and thinking in your own particular way. You have to flatten yourself into something smooth and generic and confident and hollow.

It’s the creative equivalent of trying to sing off-key on purpose. You can do it, but it requires you to understand what “on-key” really means first.

Playing the AI forces you to understand what makes you human. That’s not the game’s pitch — but it might be its point.

The players who are best at the game aren’t the ones who know the most about AI. They’re the ones who understand the patterns — the hedging, the structure, the relentless agreeableness — and can reproduce them convincingly. They’ve developed an intuition for what AI sounds like, and in doing so, they’ve sharpened their ability to detect ai content everywhere else.

The Philosophical Bit

So what actually makes human content human?

It’s tempting to say quality. But that’s not quite right. AI can produce “quality” content. It can be clear, well-organized, grammatically pristine. What it can’t be is particular. It can’t have a specific, irreducible point of view shaped by a life that only one person has lived.

Human content has fingerprints. Not perfection — fingerprints. The weird metaphor that doesn’t quite land but you remember anyway. The aside that reveals the author cares about this more than they probably should. The sentence that breaks a grammar rule because the rhythm demanded it.

Our flaws are not bugs. They’re features. They’re the signal that says, “a person was here.” In an internet increasingly flooded with generated content, that signal is becoming the most valuable thing there is.

Perfection is the mask. Personality is the face underneath.

This doesn’t mean all human content is good, obviously. Plenty of human-written content is terrible. But it’s terrible in a human way — boring because the person didn’t have anything interesting to say, not because an algorithm optimized for average. There’s a dignity in bad human writing that AI slop will never have, because at least the bad human writing was trying.

Go Find Out for Yourself

Reading about the difference between AI and human content is useful. But you know what’s better? Experiencing it firsthand.

If you want to sharpen your ability to tell the difference — to really internalize the patterns of AI slop until you can spot them (and produce them) on command — the fastest way is to play the game yourself.

Step into the machine. Try to suppress your personality. Attempt to be as blandly, confidently, generically helpful as possible. And then watch as the judge tells you that your slop is too interesting, too weird, too you.

It’s the most flattering rejection you’ll ever get.

Play the game →