A Kapwing study published in June reviewed the first 500 videos TikTok shows a newly created account. 59% was what’s now known as AI slop: machine-generated, generic content made to rack up views without saying anything.
On YouTube Shorts the figure was 21%. Three times less.
The methodological detail matters: that 59% is what the algorithm decides to show someone it doesn’t yet know anything about. Without preference signals, TikTok falls back on whatever is abundant. And what’s abundant is content that’s cheap to produce.
Merriam-Webster chose slop as its 2025 word of the year. It’s not a niche exaggeration.
Why this is your problem and not TikTok’s
Your brand isn’t competing against other brands for that person’s attention. It’s competing against six out of ten automated videos on the same screen.
And the risk isn’t that they beat you. It’s worse: it’s that they lump you in with them.
When someone spends thirty seconds scrolling through synthetic content, they train their eye. They learn to recognize the rhythm, the type of image, the tone of voice with no person behind it. And when your post shows up (made with care, but with the same tools and the same templates) they put it in the same bag and keep scrolling.
It doesn’t get lost for being bad. It gets lost for being indistinguishable.
The irony of saying this from here
We’re an AI-based lab. We use these tools every day and we’re not going to pretend otherwise.
And that’s exactly why the distinction seems urgent to us, because it’s what determines whether this helps or does harm: AI is great at producing and terrible at deciding.
A model can write you twenty versions of a text in a minute. None of those twenty will tell you which one is best, or whether it’s worth publishing something that week, or whether the angle is honest about what your company really does. That’s still yours, and it’s the one thing that can’t be automated.
What produces slop isn’t the tool. It’s skipping the deciding part.
Five ways not to be mistaken for the pile
Have something to say before you open the tool. If you start by asking the model for ideas, you’ll end up with the same ideas as everyone who did the same. The point of view comes first.
Use material only you have. Your numbers, your cases, your mistakes. A figure from your own operation can’t be invented by any model, and it’s what makes a post impossible to replicate.
Show your hand. Behind the scenes, half-finished processes, decisions that went wrong. It’s the hardest thing to automate because it requires having been there.
Disclose the tool. Saying that you use AI, and for what, is becoming a differentiator rather than an embarrassment. In an environment where people suspect everything synthetic that isn’t disclosed, transparency builds credibility. And it sounds better coming from you than from someone discovering it.
Change the metric. TikTok’s algorithm rewards retention, but quality signals (saves, shares, genuine comments) weigh more than views. Posting twice as much won’t save you. Posting something someone saves will.
The bigger picture
The strategy of flooding the channel with automated content made sense when few were doing it. Now that almost everyone does it, volume has stopped being an advantage: it’s the background noise.
What’s scarce, once again, is having something to say.