The Only Signal AI Can't Fake on LinkedIn Is Specificity
AI writes the posts and AI writes the comments. The one thing the loop can't produce is specificity, and it's the only edge a real creator has left on the feed.
I read a comment on a LinkedIn post this morning that said, “Really valuable perspective, especially point 2.” It had eleven likes. I’m fairly sure no human wrote that comment, and I’m fairly sure no human wrote point 2 either.
That’s the feed now. AI writes the post. An AI tool reads the post and writes an approving comment. Another tool likes the comment. The whole thing loops through an algorithm that can’t tell a real observation from a pattern that looks like one, and rewards both about equally.
The usual complaint here is that it’s fake, dishonest, that it devalues the platform. I don’t think that’s the real problem, and the authenticity argument never quite held anyway. Writing has always been shaped by editors, templates, ghostwriters, tools. The line between “my voice” and “influenced by other people’s patterns” was blurry long before Claude could write a hook.
The actual problem is quieter. When AI reads AI, the loop optimizes for content that looks like content, and drifts away from the thing content is supposed to be about: something that actually happened to someone.
How the loop tightens
Someone generates a post about AI productivity tips. Clean grammar, formatted for the algorithm, moderate reach because it fits the pattern the algorithm already recognizes. A comment tool reads it and mirrors the energy back: “Great insight, this really resonated.” That comment gets a like, maybe a reply, also generated. The original poster sees engagement and writes another in the same shape. Some tool logs it as “what works.”
Over a few months the posts get more optimized for the pattern and less connected to anything that happened to a real person. The feed fills up with fluent nothing.
The one thing the loop can’t produce
Specificity. That’s the signal the algorithm can’t measure but readers can still feel.
An AI post about “5 lessons from my PM career” can be perfectly written and completely empty. Compare that to: a hiring manager once asked me, in the room, to improve the spend-categorization feature on their card product, and I had to answer on the spot. There is a specific company, a specific feature, a specific moment that exists in exactly one person’s memory. You cannot pattern-match your way into it, because it didn’t come from a pattern. It came from being there.
That’s the whole game now. The bar for generic content is zero. Anyone can produce it in seconds. The bar for specific, experience-based content is exactly as high as it always was, because you have to have actually had the experience.
Where to put the effort
Not in hook structure. Not in posting time. Not in hashtag research. Those all optimize for the layer that AI already won.
Put the effort into doing things worth writing about, and then writing them down with enough detail that the piece could only have come from you. Ship a product and watch it break in front of paying users. Sit through the interview you’re underqualified for. Build the thing you wished existed, like I did with Astrika and Jobtune, and write about what actually happened instead of what should have.
The AI content loop will keep producing volume. That won’t stop. But the readers who matter keep filtering for the real thing. They always did. It’s just harder to find now, which, if you have real things to say, is quietly good news for you.