LinkedIn Taught Us to Write Like Machines. Now It's Fighting the Result.

When WeChat started testing an AI writing helper for its Moments feed, I felt a pang of something. Not surprise, exactly—more like resignation. WeChat has always kept the plain text post hidden behind a little button, as if they knew that writing your own thoughts was something you'd do only if you really wanted to. Now they're making it even easier to let a machine do it for you.

But Moments is one of the last places on the internet that still feels human. You see your friend's blurry sunset photo, the one with the overexposed sky, and their half-baked thought about whether their cat is actually plotting something. It's messy, but it's real. If AI starts ghostwriting those posts, what's the point? I think the answer is already visible on LinkedIn, the world's biggest professional network.

The Floodgates Opened

Remember when ChatGPT came out in late 2022? LinkedIn was all over it. Within months they had Collaborative Articles, where you could add your two cents to an AI-generated prompt and get a little badge. Then came AI-written job descriptions, AI-assisted resumes, and a button that would rewrite your post for you. Paying users got Copilot, Microsoft's AI, built right in.

It didn't take long for the feed to fill with sludge. In July, a company called Pangram looked at about a million posts across LinkedIn, X, Medium, Reddit, and Substack. Their finding: LinkedIn was the most AI-saturated platform. 41% of long-form posts and 30% of short posts were fully machine-generated. And get this—LinkedIn alone accounted for two-thirds of all AI posts in the entire study.

What stood out to me was the split. Only 4.3% of long posts were AI-assisted. The rest were either fully human or fully AI. No middle ground. That's not an accident. LinkedIn's design pushed you toward that one-click 'Rewrite with AI' button instead of giving you gentle suggestions to improve your own draft. The message was loud: your words probably aren't good enough.

LinkedIn's Half-Hearted Backpedal

By late July, LinkedIn had seen enough. They introduced a 'Seems like AI slop' report button. They quietly removed the 'Rewrite with AI' feature, replacing it with a basic proofreader. Executives started saying that reducing AI slop was a top priority.

Here's the irony: when Pangram analyzed LinkedIn's own announcement about fighting AI content, they found it was likely AI-generated. That's perfect. When a platform spends years training millions of users to write in a certain corporate voice, of course the official statement sounds like a robot wrote it.

I've seen this pattern before. In my own feeds, I've caught myself drafting a post, then thinking, 'This sounds too much like the usual LinkedIn fluff.' But the algorithm seems to reward that fluff. So I end up trimming the edges, adding a personal anecdote, trying to sound less like a press release. It's exhausting.

We All Learned to Speak 'LinkedIn'

This isn't just a LinkedIn problem. Every platform has its own dialect. On podcasts, you're supposed to turn a minor inconvenience into a life lesson, peppered with jargon. Travel bloggers either gush with fake enthusiasm or channel a bored aristocrat while reciting Wikipedia facts. The corporate buzzwords mocked in the Chinese movie 'Annual Meeting Can't Stop' are just the tip of the iceberg.

Nobody actually likes these tropes. But we all think everyone else likes them. It's a classic case of pluralistic ignorance—we're all playing along with a consensus that doesn't exist. When everyone believes 'this is what works,' AI becomes the perfect tool to produce the most generic, safest content possible.

I remember a few months ago, I wrote a post about a project failure. I spent hours on it, trying to make it resonate. A friend of mine, who's a content strategist, said, 'Just add a line about the lesson you learned, and put it all in short paragraphs. That's what gets engagement.' I hated that she was right.

The Algorithm's Role in the Feedback Loop

Algorithms amplify this. Take the 'broetry' style that took over LinkedIn around 2019: one sentence per line, a hook that makes you stop, a sappy moral at the end, and a personal story in the middle. It's a formula. And because each line ends with a line break, people click 'see more' to expand the post. That's engagement, so the algorithm boosts it. More people see it, more people imitate it. Before long, a ridiculous format becomes the norm.

I've fallen into that trap myself. I once wrote a post in that style, got a hundred likes, and felt dirty about it. But the algorithm doesn't care about my pride.

The Real Problem: The Metrics That Used to Work Are Broken

Underneath all this is a structural issue. For a long time, polished writing was a decent signal of quality. If someone wrote a well-structured, jargon-free, lengthy post, you could assume they knew their stuff. That's why LinkedIn rewarded those features—they indicated effort and expertise.

AI destroys that signal. Now anyone can produce a 1,500-word 'deep dive' in ten minutes, and it looks perfect but says nothing. The old proxy metrics—length, structure, vocabulary—are meaningless. Finding good content is like hunting for a needle in a haystack, and the haystack just got a lot bigger.

Why We're So Angry About AI Content

It's not the AI content itself that bothers me. I chat with ChatGPT all the time, and I don't complain about its tone. The frustration comes from broken expectations. When you're scrolling and you tap on an interesting-looking post, only to realize it's an AI-generated snoozefest, the gap between what you expected and what you get is instantly irritating.

Using AI is rational for the individual. It saves time, boosts engagement odds, and keeps you on the treadmill. But if everyone does it, the commons get ruined. We save writing time but lose reading time, sifting through more nonsense to find the few real gems. Classic tragedy of the commons, and nobody wins.

What Can Platforms Actually Do?

LinkedIn's approach—letting users report AI slop—might work on a professional network where people use real names and care about their reputation. But on anonymous platforms, that could backfire. 'This is AI slop' could become a lazy way to dismiss someone's opinion without engaging. And our ability to detect AI text by eye is notoriously unreliable, as a study at Stanford showed.

The bigger lesson is that in an age of infinite AI-generated text, the definition of scarcity has changed. What's rare isn't well-formed sentences—it's genuine human voice. It's the awkward, imperfect, personal post that could only have come from a real person with a real perspective. As someone online put it, I'd rather read one honest sentence than a hundred beautiful ones.

Platforms like LinkedIn, WeChat, and others are now trying to walk a tightrope: they want AI's efficiency without killing the goose that lays the golden egg—authentic human contribution. It's a hard balance, and the jury's still out. But one thing is clear: the value of being human is only going up.