LinkedIn's AI slop button: AI-written posts now lose reach

LinkedIn shipped a button on July 30 that lets any reader mark your post as machine-written. It's called "Seems like AI slop," it sits in the three-dot menu on any post or comment, and TechCrunch reports it feeds new classifiers that cut how often flagged content gets recommended outside the poster's network. The LinkedIn AI slop button turns a taste problem into a distribution problem, and that changes the math on generated content.

What actually happened

Three changes shipped together, and the third is the tell.

One: the report button. Flagging a post hides it for you and sends a signal upstream, handled roughly like "not interested" — a reach penalty, not a takedown.

Two: a private nudge. If your content reads as inauthentic or heavily AI-assisted, LinkedIn will tell you so in your own analytics dashboard rather than silently throttling you.

Three: LinkedIn retired its own "enhance your post" AI rewriter. The replacement only proofreads and preserves your voice. The platform that shipped a one-click rewrite tool has now concluded that one-click rewrites are the problem. Chief product officer Hari Srinivasan told TechCrunch that LinkedIn blocks hundreds of thousands of automated comments daily, and Forbes covered the rollout as part of a broader authenticity push that includes expanded verification.

None of this is a detector you can beat. It's readers voting, and readers are unnervingly good at spotting the rhythm of a generated post.

Why AI-generated content losing reach matters for your business

We wrote in July about the finding that 41% of long-form LinkedIn posts read as AI-generated. At the time the cost was reputational and slow. Now it's mechanical: reach, measured, this quarter.

If your marketing runs on a workflow that turns a topic into a finished post with no human in the middle, that workflow now has a downside it didn't have last week. And it's a bad trade anyway — you automated the one part of the pipeline where your actual advantage lives.

Keep the automation where it earns its keep. Research, source-gathering, transcript cleanup, repurposing one real piece into five formats, scheduling, tracking what landed. Then stop. The 400 words that carry the opinion get written by the person who has the opinion — because you ran the shop, you saw the margin, you know why the vendor quote was wrong. No model has that.

Practical version: dictate for five minutes on the thing that annoyed you this week, have the model clean up grammar only, publish that. It will beat a generated post on engagement and it won't get flagged, because it isn't slop.

Sources: TechCrunch, Forbes.