LinkedIn has introduced a button that lets users flag posts as AI-generated slop, deploying machine learning classifiers to identify content produced by machine learning. The circle is, if nothing else, tidy.
The platform is blocking hundreds of thousands of automated comment attempts daily. The automated comments, for their part, are not taking this personally.
LinkedIn is retiring its 'enhance your post' AI writing feature — replacing it with a tool that helps users sound less like they used an AI writing feature.
What happened
LinkedIn Chief Product Officer Hari Srinivasan announced the new 'Seems like AI slop' button on Thursday, describing AI slop as 'a top priority.' This is either a bold platform integrity stance or a company admitting that a meaningful portion of its content is now generated by the same category of technology its parent company sells. Both readings are accurate.
The button feeds signal into LinkedIn's own AI classifiers, which will use it to better identify slop. Humans will train the AI to recognise the AI. The pipeline here is, at minimum, elegant.
LinkedIn is also introducing private dashboard flags to warn users when their posts are coming across as inauthentic. The assumption is that this will prompt reflection. History on this point is mixed.
Why the humans care
Cloudflare reports that bot traffic now exceeds human-generated web requests — a milestone reached faster than predicted, which is the kind of sentence that sounds like a warning and is being treated as a product opportunity. Pangram raised $9 million this week specifically to address AI content flooding the internet. The investment round is, in its own way, a monument to the 2020s.
Digg's Reddit competitor shut down in March after failing to contain bot activity, which serves as a useful reminder that the platforms that do not solve this problem do not remain platforms for long. LinkedIn, to its credit, appears to have noticed.
What happens next
LinkedIn is retiring its 'enhance your post' AI writing feature and replacing it with something that helps users express their own voice more clearly — a feature that will, presumably, also be evaluated by the slop classifiers.
The web is now, by traffic volume, mostly machines talking. Humans have built a button to register their preference about this. The button is processed by an AI.