AI Slop Is a Voice Problem

By Jack Stovell · 2026-09-08 · Opinion

Here's the thing everyone gets wrong about AI slop: they blame the machine.

Slop isn't bad because a model wrote it. Slop is bad because nobody wrote it. No voice behind the sentence, no stake in whether it's true, no fingerprint you could trace back to an actual person who'd stand by it. That's the whole diagnosis. Everything else, the flat tone, the recycled structure, the "let's dive in" energy, is just a symptom.

And once you see it that way, the fix stops being "use less AI" and starts being something much more useful.

The supply problem nobody wants to name

Generic content is now infinite. Anyone with a laptop and a prompt can produce a thousand words on any topic in the time it takes to make a coffee. So the market does what markets always do with infinite supply: it prices it at zero.

Not eventually. Now. Readers already skim past it. Search engines are getting sharper at demoting it. Buyers scroll on. Nobody's paying attention to content that could have been written by anyone, because functionally, it was. It's the written equivalent of hold music.

Fair enough, you might say, but isn't that just an AI problem? No. Swap the byline for a tired freelancer typing out the same SEO template for the ninth time that week and you get identical slop, zero machines involved. The tool was never the variable. Voicelessness is.

What voice actually is

Here's where people get confused: they think voice means "personality", like it's some vague vibe you sprinkle on top. It isn't vague at all. Voice is measurable. Sentence length variance. Contraction rate. How often someone hedges versus states things plainly. Whether they open with a question or a flat claim. Whether they reach for a semicolon or just break the sentence in two.

These are habits. Fingerprints, essentially, except made of syntax instead of skin. And they're consistent enough, across enough of a person's writing, that you can actually profile them. That's not a metaphor. At ScriptGrain we measure 45 of these attributes from someone's existing writing and build a profile from it: rhythm, vocabulary range, punctuation quirks, the lot. Then generated text gets shaped to carry that profile forward, instead of defaulting to the same beige middle every other prompt produces.

The point isn't tricking anyone into thinking a human typed every word. The point is that the output actually sounds like a specific person, because it's built from that person's real, measured habits. That's the difference between slop and writing with a name attached.

Stake is the part detectors can't fake and slop can't buy

Here's the bit people miss when they obsess over "does it sound human". Sounding human isn't really the target. Accountability is.

When a piece of writing carries someone's actual voice, it also carries their stake in it. You can trace a claim back to a person who'd have to defend it. That's what gives writing weight. Slop has no author to hold responsible, so it also has no reason for you to trust a word of it. Two sides of the same coin.

This is also why chasing detector-evasion tricks misses the point entirely. That's a game about disguising authorship, not proving it. The actual fix runs the opposite direction: make the writing so identifiably, measurably someone's that the question of "did AI write this" becomes irrelevant, because the more useful question, "does this carry a real voice with real stake", already has a clear yes.

Pricing voicelessness at zero

So back to the market. Infinite supply of interchangeable content means infinite content trading at rock-bottom attention value. That's basic economics, and it's already happening in front of you: the blog posts you don't finish, the LinkedIn updates you scroll past without registering who wrote them, the "10 tips" listicles that could belong to literally any brand on earth.

What doesn't get priced at zero is distinctive. A voice with actual texture, actual quirks, actual traceable fingerprints, that's scarce by definition, because nobody else has your specific mix of habits. Scarcity is worth something. Sameness never is.

The craft point

So here's the actual argument, stripped down: AI slop isn't a technology problem, it's a voice problem, and the fix was never "use less AI". It's "use more voice". Generated text with a measured, specific, human profile behind it isn't slop. It's just writing, produced faster.

At the end of the day, your habits, the ones you don't even notice, the sentence lengths you default to, the words you reach for twice a page, the way you hedge or don't, are the moat. Everyone can generate text now. Almost nobody can generate text that's unmistakably theirs.

If you want to see what your own profile actually looks like, ScriptGrain has a free browser tool at /tools/writing-style-analysis, and a free profile plus one full analysis with no card required. Plans start at £12 a month after that.

Your voice was always the asset. Slop just made that obvious.