LinkedIn Says Its Anti-AI Slop Button Actually Worked

[Freddie_85]

LinkedIn's chief product officer Hari Srinivasan has shared an update on the platform's Seems Like AI Slop button, saying it has already been used more than a million times in its first two weeks of existence, and that posts flagged by the tool are now getting roughly forty percent fewer views than similar content was getting just a few weeks earlier. The button itself sits inside the three dot menu on any post or comment, showing a small diamond icon next to the phrase Seems like AI slop when a user taps it.

The scale of the underlying problem the button is meant to address is genuinely enormous. Independent research from AI detection company Pangram previously found that around forty percent of long form posts on LinkedIn, meaning anything over 250 words, were likely fully AI generated, giving the platform the highest rate of any major social network the company measured. LinkedIn was also simultaneously encouraging exactly this kind of content for years through its own enhance your post feature, an AI powered rewriting tool that has now been quietly killed off and replaced with a more limited proofreading assistant that claims to preserve a person's original voice rather than rewriting it wholesale.

LinkedIn is also rolling out a new notification feature tied into post analytics that will privately alert someone when other users have flagged their specific post as AI slop, giving posters direct feedback on how their writing is actually landing rather than leaving them to guess. Srinivasan framed this as training a person to sound more authentic based on real human judgment rather than simply running their draft through an automated AI detector that could easily get the call wrong.

Critics have already raised a fair concern about exactly what this crowdsourced flagging system is actually teaching LinkedIn's underlying models to detect. Since most casual users genuinely cannot reliably distinguish AI writing from a human who simply writes in a clean, polished, corporate style, the button risks training LinkedIn's classifiers to recognize suspicion and personal bias just as much as it trains them to recognize actual machine authorship, especially once workplace rivalries and simple dislike of a particular poster start factoring into who gets flagged.

Whether a forty percent drop in views for flagged content actually reflects better detection or simply reflects users collectively piling onto whatever already looks unpopular is a genuinely open question that Linkedin's own success metrics do not yet seem equipped to answer

Linkedin Says Users Love Its Anti Ai Slop Buttonwww.pcmag.com