Back in spring the platform started throttling what it calls 'AI slop'. Low-effort, machine-made posts don't get deleted, they just stop moving. They reach your direct connections and die. By July you could flag a post with a "Seems like AI slop" button, and the feed folds your verdict into its training.
Everyone read that as LinkedIn declaring war on AI. It wasn't. And the real version is a bit funnier.
LinkedIn's been weirdly careful with the words. Their line is that AI and slop aren't the same thing. The classifier doesn't care whether a model wrote your post. It cares whether the post is generic, hollow, polished on top and empty underneath. You can produce that with ChatGPT in nine seconds. You can also produce it by hand, and people have, for years, long before the robots showed up. One write-up nailed it: the penalty is for being boring, not for being AI.
Sit with that for a second, because it stings in a way the AI version doesn't.
"I got flagged for using ChatGPT" has an escape hatch. Everyone's doing it, the tool wrote it, not really me. But "I got flagged for being boring"? There's no hiding in that one. The machine read what you wrote, went looking for a single original thought, found the shape of one with nothing inside it, and decided the rest of the room didn't need to hear from you. You can blame a tool for a bad sentence. You can't blame it for having nothing to say.
And then the number that actually got me. Someone ran a detector over long-form posts from a pile of big, influential LinkedIn accounts, and more than half, around 54%, came back as likely machine-written. More than half the thought leadership on the professional world's main stage, and nobody's home in any of it. These aren't junk accounts. These are the people selling you their personal brand, their expertise, their voice. Half of them outsourced the voice - not quite sure if that makes me want to laugh or cry.
The painfully ironic detail, unbearably funny, in a bleak way; some of the most AI-sounding posts on the entire platform are AI-written complaints about AI slop. People prompting a model to manufacture outrage about people prompting models. A feed slowly eating its own tail and calling it engagement.
So what do you do if you're a business trying to be seen there? Not stop using AI, that ship's gone and LinkedIn isn't even asking. The lesson's smaller and more uncomfortable than that. The feed stopped rewarding how much you post and started grading what's actually in it, and the only thing it's really checking for is whether a real person with a real view turned up. A claim someone could argue with. A number you went and measured. An opinion that cost you something. The posts it buries are the ones that could've been written by anyone, about anything, for no one.
None of which is new. It's the oldest rule there is. Say something worth reading. The only thing that changed is a platform finally bolted a punishment onto it and flipped the switch across half the feed overnight.
The machines got good enough to spot writing with no human left in it. So the rarest, most valuable thing you can be right now is a person with something to say. It's a bit sad that it took an algorithm to remind us. But here we are, and the ones who still have a voice are about to look very, very expensive.