LinkedIn is grading your posts now.

LinkedIn has added a reporting option that lets users flag a post as AI slop, available from the three-dot menu on any post in the feed. Chief Product Officer Hari Srinivasan announced it on Thursday, July 30, alongside a set of upgraded classifiers built to identify machine-written and low-quality content. Flagged material will appear less often in suggested content and in posts from outside a user’s network, and every report becomes training data for the classifiers.  

To understand why the platform bothered, start with the number that prompted it. AI detection company Pangram scanned more than a million posts across LinkedIn, Medium, Reddit, Substack, and X, and found LinkedIn carried the most long-form machine-written content of any platform measured, with more than 40% of posts longer than 250 words reading as fully AI-generated. Pangram chief executive Max Spero has described that content as a tax on readers’ time. Srinivasan’s framing was less clinical. “AI slop is a top priority for all of us,” he wrote.  

The part that got less attention is the more interesting one. LinkedIn is retiring “Enhance your post,” its own AI writing tool, and replacing it with a narrower proofreading feature intended to preserve the writer’s original voice. The company will also privately notify people when their content reads as inauthentic because of heavy AI use.  

Read that sequence in order. LinkedIn built a tool that helped produce the content, watched what the feed became, and switched the tool off. There is no version of that decision that isn’t an admission. A platform whose entire product is professional attention looked at the tradeoff between volume and readability and picked readability, which is not the choice most software companies made over the last three years. 

This is also not an isolated move. Substack launched its own Pangram-built detection feature on July 21, letting readers scan any post, note, comment, or reply over 100 words for an estimated split between human and machine writing, with the option for writers to add a disclosure, turn detection off on individual posts, or dispute a result. Two platforms with different business models arrived at the same conclusion within ten days. The backdrop is Cloudflare’s finding that automated bot traffic on the web now exceeds human traffic, a threshold it crossed earlier than expected.  

So what does this mean if LinkedIn isn’t where you spend your time? The mechanics of distribution just changed. Reach beyond your existing network is now conditional on passing a classifier, which means the cost of publishing something generic went from low to negative. For years the safe play in B2B was to post consistently and let volume do the work, because the algorithm rewarded frequency and nothing punished blandness. Something now punishes blandness, and it runs automatically on every post. 

There is a second signal here worth reading, and it arrived before LinkedIn did. Analysis from Originality.ai, a competing detection firm, found that AI-generated LinkedIn posts drew an average of 45% less engagement than human-authored posts across its dataset. The audience had already discounted this content without needing a label or a button. The platform is catching up to a judgment its users made on their own, which is the part that should reframe how you read the announcement. This is not a policy imposed on the feed. It is the feed’s existing behavior, made into a product.  

One caution before anyone treats detection as settled. These scores are estimates built on stylistic patterns rather than proof of how text was produced, and detection specialists have warned that false positives fall hardest on formal, predictable, or non-native English writing, where a human corporate announcement can read like a machine wrote it. Spero himself has called Pangram’s figures a lower bound, since people who install a detection extension are not a random sample. Any B2B team with contributors writing in a second language should expect noise, and there is currently no published appeals process on the LinkedIn side. 

For B2B marketers the practical takeaway is narrower than the headlines suggest. The tools are not the exposure. Editing a draft, tightening a structure, or fixing a sentence leaves nothing for a classifier to find. What gets flagged is the post with no observation in it, the one assembled from a prompt because something had to go out on Tuesday. That post was always weak. It just used to travel anyway. 

The classifier isn’t the threat. Having nothing to say is.