If you spend any time scrolling through LinkedIn, you may have noticed that many posts sound oddly similar: generic motivational platitudes, corporate buzzwords, and perfectly structured paragraphs that feel like no human actually wrote them. That is because a large portion of what appears in the feed is generated by artificial intelligence. To address the growing flood of automated content, LinkedIn is rolling out a dedicated “Seems like AI slop” button, allowing users to flag posts and comments they believe were produced by AI.

The announcement was made by LinkedIn’s chief product officer, Hari Srinivasan, on the platform itself. “We are ramping the ability for members to tell us if they believe a post or comment seems like AI slop,” Srinivasan wrote. “Slop is hard to define and the definition changes; this lets us tune our models and make better feeds.” The feature is part of a broader push by LinkedIn to reduce low-quality, machine-generated content and improve the overall user experience.

Key facts

• LinkedIn is adding a “Seems like AI slop” reporting option accessible via the three-dot menu on posts and comments.
• Users whose posts are flagged as AI slop will be privately notified through their analytics dashboard.
• LinkedIn says it has already blocked billions of AI-generated posts and comments in the last couple of months.
• Substack is integrating the AI detection tool Pangram, and its CEO has publicly criticized LinkedIn’s AI content problem.
• Pangram’s data suggests that as much as 40% of long-form text on LinkedIn is AI-generated, and that LinkedIn accounts for 62% of all AI content flagged by the tool.
• LinkedIn will retire its “enhance your post” feature and replace it with a proofreading tool designed to preserve the user’s original voice.

How the new reporting button works

The new feature is designed to be simple. On any post or comment, users can click the three-dot menu in the upper-right corner and select the button labeled “Seems like AI slop.” Once a post is flagged, the person who shared it will receive a private notification in their analytics dashboard. That means users whose content is flagged will be made aware that others found it to be inauthentic, but the system does not appear to be a public shaming mechanism. Instead, LinkedIn is using the feedback to fine-tune its machine-learning models and adjust how content is ranked in feeds.

Srinivasan stressed that the definition of slop is not static. “Slop is hard to define and the definition changes,” he said. This acknowledgment is significant because AI-generated content is constantly evolving. What looks obviously machine-written today may become more sophisticated tomorrow, making detection an ongoing cat-and-mouse game. The new button gives LinkedIn a human-in-the-loop signal that can be used to train its classifiers.

The scale of LinkedIn’s AI slop problem

The move comes as AI-generated content has become a major issue across social media platforms. LinkedIn, in particular, has become a magnet for automated posts that mimic the platform’s characteristic professional tone. Many users have complained about seeing the same kinds of formulaic posts over and over: stories about overcoming adversity, humble brags about career milestones, and motivational quotes accompanied by stock images.

Srinivasan said that LinkedIn has already blocked billions of AI-generated posts and comments in just the last couple of months alone. That staggering number underscores how pervasive the problem has become. But LinkedIn is not the only platform struggling with AI slop. Substack, the newsletter platform, has also been dealing with an influx of machine-generated writing. Earlier in July, Substack CEO Chris Best announced that the company was integrating Pangram, an AI detection tool, into its platform. Best did not mince words when explaining the decision.

“We built the ability to do a Pangram scan into the Substack app, because we’re sick of slop and we don’t want substack to turn into LinkedIn,” Best wrote in a series of posts on X. “Platforms that reward fakeness will drive a race to the bottom. Pangram’s data shows that as much 40% of long-form text on LinkedIn is generated.” That comment was widely interpreted as a direct shot at LinkedIn’s reputation as a hub for low-effort, AI-assisted corporate content.

Pangram’s own blog post from earlier in the month supported Best’s claim. According to the company’s data, LinkedIn has the most long-form AI content out of all popular social media platforms that host long-form content. “LinkedIn posts made up a third of scanned items, yet it accounted for nearly two-thirds (62%) of all AI content we flagged,” Pangram said in the post. These figures highlight just how much automated content is circulating on the platform, despite LinkedIn’s previous attempts to curb it.

Srinivasan’s response to critics

While Srinivasan’s announcement did not directly attack Substack in the way that Best attacked LinkedIn, it contained what many observers saw as a subtle jab at the use of automated detection tools. “We want members to get feedback from real humans on what sounds authentic – not just have an AI detector review it and get it wrong,” Srinivasan wrote.

This comment touches on a known problem with AI detection software. These tools are not perfect. They frequently produce false positives, flagging human-written content as machine-generated. Many AI detectors rely on statistical patterns and perplexity scores, which can be thrown off by formal language, technical jargon, or even well-structured prose. Moreover, many AI detection tools are themselves powered by AI, making it difficult for users to understand why a particular piece of text was flagged. The reasoning behind the verdict is often hidden in a black box, leaving writers with no clear path to appeal or correction.

There are also concerns that AI detection tools discriminate against non-native English speakers. Writers who use straightforward grammar, avoid idiomatic expressions, or follow rules of style more rigidly may be more likely to be classified as AI-generated. Studies have shown that humans themselves are not much better at identifying AI text. Previous research has characterized people’s ability to distinguish between human and AI writing as little better than a coin toss. This is why LinkedIn’s approach of relying on human feedback is interesting, even if it is imperfect.

Broader changes to LinkedIn’s content system

In addition to the new reporting button, Srinivasan announced that LinkedIn would be “ramping up a series of new and improved classifiers that identify if a post is AI-slop or generally low-quality content.” These classifiers will work alongside the human reporting data to identify problematic content more accurately and remove it from feeds.

LinkedIn is also making changes to one of its AI-powered writing tools. The platform will be retiring its “enhance your post” feature, which users could use to rewrite their drafts in a more polished or professional tone. This feature was notoriously easy to abuse: users could write a rough outline and let the AI turn it into a complete, polished post, often with generic corporate language. The replacement is an AI tool that proofreads content but keeps the user’s original voice intact. This is a notable shift, as it suggests LinkedIn wants to use AI to assist with grammar and clarity without stripping away the personality that makes content feel human.

The challenge of defining slop

One of the central difficulties with the new “Seems like AI slop” button is that there is no universally accepted definition of slop. Slop is not simply content that is grammatically correct or uses certain punctuation marks. Many legitimate writers use em dashes, colons, and other stylistic devices that are also common in AI-generated text. The term “slop” itself is subjective, and what one user considers authentic may be seen as low-quality by another.

There is also the question of where to draw the line. Is a post that was lightly edited by AI considered slop? What about a post that was outlined by AI but written by a human? Or a post that uses AI for brainstorming but not for the final text? These gray areas make it difficult to enforce any strict policy. LinkedIn’s approach appears to be to gather as much feedback as possible and let its models learn from the aggregated signals, even if individual judgments are inconsistent.

The platform has a long history of trying to balance professional networking with engagement-driven content. In recent years, LinkedIn has become more social and more consumer-like, with users sharing personal stories, opinions, and even memes. This shift has made the platform more lively, but it has also made it more susceptible to the kinds of viral, low-effort posts that AI tools are particularly good at producing. The result, many users feel, is a feed that often resembles a mix of endless self-help advice and corporate announcements rather than genuine professional insights.

What the future holds for AI detection and social media

LinkedIn’s decision to add a human reporting button is just one example of how social media platforms are adapting to the rise of generative AI. Many companies are experimenting with labels, machine-learning classifiers, and community reporting systems to manage the growing amount of synthetic content. Substack’s integration of Pangram is another example, though it relies on automated detection rather than human feedback. Both approaches have strengths and weaknesses, and it is likely that the most effective strategies will combine multiple signals.

The broader issue is not limited to LinkedIn or Substack. Social media platforms such as Facebook, Instagram, X, YouTube, and TikTok are all dealing with AI-generated text, images, and videos. Some platforms have introduced disclosure requirements, while others have focused on removing spam and inauthentic accounts. But the speed at which generative AI is improving means that platforms are constantly playing catch-up. What is detected today may be undetectable tomorrow, and the tools that can identify current versions of AI text may quickly become obsolete.

There is also a commercial dimension to this fight. AI-generated content is often created to drive engagement, sell products, or build personal brands, and it can be very effective at doing so. Even on LinkedIn, where the audience is mainly professionals, a well-timed motivational post can generate thousands of likes and comments. This incentivizes users to continue using AI tools, despite the platform’s efforts to discourage them. Whether the new “Seems like AI slop” button will change that calculus remains to be seen.

For now, LinkedIn is moving forward with its plan to clean up the feed. The company is investing in new classifiers, retiring its most easily abused writing feature, and opening the door for users to actively participate in the moderation process. The success of these efforts will depend on how well LinkedIn can balance the desire for authentic human connection with the realities of an online ecosystem where AI is increasingly everywhere. Expect to see less slop in your LinkedIn feed, the executive promised.

Source:Gizmodo News