# LinkedIn gave everyone a 'this looks like AI' button. I won't press it. > Source: > Published: 2026-08-04 11:29:33+00:00 This week LinkedIn [rolled out](https://techcrunch.com/2026/07/30/linkedin-adds-a-button-to-report-ai-generated-slop/) a "Seems like AI slop" button. Tap it under someone's post and the post disappears from your feed, while your report goes off to train their filter. I heard about it through work — keeping an eye on what platforms are up to is part of the job — but I read it less as a marketer and more as someone who writes, and who now, apparently, walks around under suspicion. LinkedIn's logic isn't hard to follow. According to the AI detector Pangram, they have [the highest share of AI content](https://www.pangram.com/blog/ai-in-your-feed) of any platform: roughly a third of short posts and over forty percent of long ones. The feed really is drowning, and handing people a valve is a reasonable move. It isn't the valve that bothers me. It's whose hands it got put in. The "this is AI" verdict used to come from an algorithm, and even that missed plenty. Pangram itself claims a 0.01% false-positive rate — which sounds impressive until you multiply it by volume. One hundredth of a percent of millions of posts is still thousands of real people getting labelled as bots. And that's the machine, the one somebody at least tested. Now the "this is AI" stamp has been handed to every passerby, and the cost of getting it wrong is zero. Tap, post gone, on with your coffee. You'll never even find out you were wrong — there's no feedback loop pointing back at you. I haven't touched the button once, and I don't even like AI text — the smooth filler I spent [a whole post](https://dev.to/eugeniya_ivanova_4a58eadc/i-trained-an-ai-to-sound-like-me-then-spent-three-rounds-undoing-it-3k96) pulling apart last time. But disliking slop and swatting a person are different reflexes with different results. Someone posts the best thing they know how to write, and someone else decides it "looks like a bot." That's one reader gone. That isn't fighting slop. That's slapping the hands of someone who's only just reaching for the keyboard. Here's the personal part. I write with AI and I don't hide it, but the text is mine: a pile of rules, defined styles, years of tuning, every paragraph edited by hand. Slow — though it used to be slower. AI isn't my author, it's a tool I don't let grab the wheel. I wrote a whole article about that. Then something happened that made it feel a lot more personal. A while back, in a work chat, I answered a former colleague — a plain message, a couple of paragraphs. She reads it and says: "Wow, you sound just like GPT — fast and structured." A compliment. The catch is that about two minutes passed between her question and my answer. There wasn't even enough time to open ChatGPT, write a prompt, wait for an answer, and paste it back. It was just a human who's spent a long time working with words. And it read as a machine. And for a couple of days I was pleased about it — imagine, mistaken for GPT, fast and clean. Now it turns out that's exactly the thing I'm supposed to worry about. Same sentence. Same moment. A week later it had gone from a compliment to evidence, while nothing about me had changed. What changed is that there's now a button for it. Fluency is being read as forgery. Fast and clean means you didn't write it yourself, and there's now a button wired to that assumption. The irony is that AI was trained on well-edited human prose — which is exactly why it comes out so smooth and combed. And that smoothness is now how we try to spot it: we hunt for the fingerprints of a good editor, and we reliably find them on people who simply know how to write. Congratulations to all of us. My own way out of this is as old as it gets: divide and conquer. The machine is fast and a little too good — and sometimes not good at all, and that's the line I can still see and it can't. I handed it the posting across networks; works great — it's mechanical anyway. But writing is different. No matter how much I train it, it keeps sliding back into "we did it this way before, so this must be right." Except the new piece in front of it doesn't fit the old rules, and it doesn't notice that — I do. So here's the split: authorship and judgment stay with the human, logistics go to the machine. That's how I run it through [Publora](https://publora.com) (I'm on the team): I write by hand, and the delivery across platforms goes to an agent. No AI writes my posts — it's on backup duty, in the corner, not at the wheel. How this ends, honestly, I don't know. For now I've stocked up on popcorn and I'm watching from two angles. Angle one: people start writing worse on purpose — breaking the rhythm, sprinkling in typos, watering it down, anything to avoid looking suspiciously smooth. A platform that penalizes quality, and a crowd cheerfully dumbing itself down to a C. Angle two: some just stop posting, because the fun runs thin when any passerby gets to weigh your fluency for "real or not." Which one wins, I won't guess. Popcorn's ready. I still won't press the button. Slop exists, no argument, plenty of it. But pressing something in two seconds, with zero accountability, that sorts people into "human / not human" — that I'm not up for. By the way, a huge secret: I wrote this article myself, by hand. The formatting and posting I handed to an agent — the writing was mine, the publishing wasn't, and that's exactly how I want it. Would you press it? And have you ever caught yourself deciding someone's fast, clean writing was too smooth to be real?