LinkedIn introduces a 'This appears to be AI-generated content' button.

      LinkedIn has devised a label for the somewhat motivational, AI-generated posts that clutter its feed, and it has introduced a button for this purpose. The professional networking site has discreetly incorporated a “Seems like AI slop” feature in its post menu, allowing users to report content they suspect was created by a machine, according to 404 Media.

      This button is located under the three-dot menu at the top of any post. When clicked, LinkedIn acknowledges the user’s action with a message: “Thanks for letting us know. Your feedback helps improve the feed.” Following that, the post is concealed. 404 Media mentioned that it took approximately six seconds of scrolling to find an example, which stood out due to its format, use of emojis, bullet-point summaries, and the “X is not Y” structure.

      What is particularly noteworthy is the phrasing. Instead of stating “this looks AI-generated,” the button uses the term “slop,” a straightforward internet slang term for low-quality machine output, now integrated into a platform owned by Microsoft.

      Reasons Behind LinkedIn’s Action

      This decision comes after the platform's crackdown in May on generic AI posts, which aimed to limit such content from recommendations instead of removing it outright. The new button transforms this detection into a crowd-sourced effort, with each report contributing to training LinkedIn’s own models to recognize this type of content.

      Hari Srinivasan, LinkedIn’s chief product officer, verified this in a post following the report. “AI slop is a top priority for all of us,” he stated. The company is “ramping up a series of new and improved classifiers” to identify slop and enabling users to flag posts to enhance “our models and create better feeds.”

      He included an important clarification. “AI and slop are not the same thing,” Srinivasan noted. Many users refine their ideas using AI, prompting LinkedIn to experiment with a private notification in the analytics for posters when their writing “may have appeared inauthentic.” Additionally, it is replacing its “enhance your post” tool with a proofreader that maintains the user's voice.

      The Extent of the Issue

      Recent statistics highlight the urgency of the situation. Earlier this month, 404 Media reported that the detection service Pangram estimated that 41 percent of long-form LinkedIn posts and 30 percent of short posts are likely AI-generated. Searching for “LinkedIn AI” yields numerous guides, even some hosted on LinkedIn, instructing users on how to automate their posts.

      This situation presents a notable irony, as LinkedIn has its own AI writing assistant designed to draft profiles and posts for Premium users. Its parent company, Microsoft, is also one of OpenAI’s main investors, effectively selling both the tools for creating content and the means to clean it up.

      A Broader Battle Against Slop

      LinkedIn isn’t alone in this struggle. The preprint server arXiv has begun banning researchers who submit papers laden with AI content, and various platforms have tightened regulations on automated AI channels. This push reflects a digital landscape where bots now outnumber humans in terms of traffic.

      Whether the report button will have a significant impact remains to be seen. Slop, as Srinivasan acknowledged, is challenging to define and the criteria continue to evolve. However, encouraging users to identify it openly indicates a recognition that there is indeed a problem with the feed. It also illustrates that demonstrating something is human is becoming a task in its own right.