LinkedIn is taking aim at the flood of algorithmically generated content on its platform with a new reporting option labeled “Seems like AI slop.” The feature gives members a direct way to flag posts and comments that appear to be produced by AI rather than a real person. The move is part of a broader push to clean up feeds that have become crowded with generic, formulaic posts.
Hari Srinivasan, LinkedIn’s chief product officer, announced the feature in a post on the platform. “We are ramping the ability for members to tell us if they believe a post or comment seems like AI slop,” he wrote. “Slop is hard to define and the definition changes; this lets us tune our models and make better feeds.”
How the new flagging button works
Users will find the new option in the three-dot menu in the upper-right corner of any post. From there, they can select the button labeled “Seems like AI slop.” If a post is flagged, the person who shared it will be privately notified in their analytics dashboard. That means the flag is not just a behind-the-scenes signal for LinkedIn; it also gives writers and publishers a direct warning that their content may be perceived as automated or inauthentic.
The feature leans on community feedback rather than relying entirely on automated detection. This is a meaningful shift because AI-slop definitions are subjective and constantly changing. What reads as robotic to one person might seem acceptable to another. LinkedIn’s approach is to let human signals help train its recommendation systems and sorting algorithms.
A platform-wide problem
The new button arrives as AI-generated content spreads across nearly every part of the social web. LinkedIn has been particularly affected. Srinivasan said the company has already blocked billions of AI-generated posts and comments in the last couple of months alone. The scale of the problem is enormous, and it has quickly become one of the defining challenges for social platforms in the current AI boom.
Much of the content is produced by AI writing tools that generate career advice, motivational messages, and supposedly professional insights. These posts often share a similar tone: confident but vague, polished but lacking real experience. The result is a feed that can feel hollow, repetitive, and strangely disconnected from actual human conversation.
Substack takes its own shot at LinkedIn
LinkedIn is not the only platform wrestling with AI slop. Substack CEO Chris Best recently announced that Substack would integrate Pangram, an AI detection tool, into its app. Best was blunt about the motivation. “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,” he said in a chain of posts on X.
Best’s comments were a direct jab at LinkedIn’s reputation as a place where much of the long-form content appears to be generated. Pangram’s data, cited by Best, suggests that as much as 40% of long-form text on LinkedIn is AI-generated. According to the company’s findings, LinkedIn has the most long-form AI content among popular social platforms that host long-form posts.
Pangram said LinkedIn posts made up about a third of all scanned items but accounted for nearly two-thirds, or 62%, of all AI content it flagged. That is a striking imbalance. It suggests that while AI-generated content exists all over the internet, LinkedIn has become one of its strongest breeding grounds.
Srinivasan did not attack Substack as directly as Best attacked LinkedIn, but his announcement contained what looked like a subtle jab at 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,” he wrote. That line touches on a growing debate about whether AI detection tools are reliable enough to be the final word.
The limits of AI detection
Commercial AI detection tools are increasingly popular, but they are not foolproof. They can produce false positives, flagging human-written content as AI-generated. Many of these tools are themselves powered by AI models, which means their reasoning is often hidden from users. When a writer is told their work may be AI, there is no clear way to challenge the verdict or understand exactly what triggered it.
The limitations go beyond technical accuracy. AI detection tips often emphasize grammar and stylistic elements that are not unique to AI. Features like em dashes, colons, and carefully structured paragraphs are common in professional writing. Applying these rules too rigidly can unfairly target non-native English speakers, who may write in a more formal or conventional style.
Human judgment is also weak. Studies have shown that people identify AI-generated text at rates that are barely better than a coin toss. That makes the new “Seems like AI slop” button an interesting experiment: it relies on the crowd, but the crowd is not always right. Still, LinkedIn is hoping that aggregated signals from many users can be more useful than any single automated evaluation.
LinkedIn’s broader fight against low-quality posts
Along with the new flagging button, Srinivasan said LinkedIn would ramp up a series of new and improved classifiers designed to identify AI slop and generally low-quality content. These classifiers will work alongside the human reporting system to surface better material and push down posts that do not add value.
LinkedIn is also planning to retire its AI-driven “enhance your post” feature. The tool was designed to help users improve their writing, but it may have contributed to the very sameness that makes so much LinkedIn content feel automated. In its place, the company is introducing an AI tool that proofreads while keeping the writer’s voice intact. The goal is to offer editing support without flattening a person’s style into generic corporate boilerplate.
The changes reflect a larger tension in the AI era. Platforms want to use AI to help people create, but they also want to prevent AI from degrading the quality of public conversation. Writing tools can boost productivity and support people who struggle with grammar or structure. But when the same models are used by millions of people, the output starts to look the same.
For LinkedIn, the challenge is particularly difficult because the platform rewards professional-sounding language. Users are encouraged to share insights, celebrate achievements, and offer advice. That environment is ripe for automation. AI can easily produce a post about leadership lessons or career growth, and it takes effort to tell the difference between a genuine personal story and a smoothly generated one.
The new flagging option gives users a small amount of power over their own feeds. It also creates a feedback loop that could help LinkedIn understand what its members consider authentic. Whether that leads to a noticeably cleaner feed is still an open question, but the company is clearly betting that human input will help it do a better job than relying on AI alone.
The private notification for flagged users is an interesting choice. In many moderation systems, users are not told when their content is flagged. LinkedIn’s approach offers a degree of transparency, but it also creates a potential for friction. A person who spent time writing a thoughtful post may be surprised to learn that their content is considered AI slop, especially if the flag is incorrect.
The distinction between AI-generated content and AI-assisted content is also blurry. Many professionals now use AI tools to outline, edit, brainstorm, or refine their writing. A post might be drafted by a human and polished by AI. Is that slop? The answer likely depends on how much AI was involved and whether the final product reflects genuine experience. That is another reason human detection is so tricky.
LinkedIn’s effort also comes amid broader societal concerns about AI-generated content. Fake reviews, spam comments, and automated news articles have already eroded trust in online information. The term “slop” has become a shorthand for low-quality, mass-produced media. Platforms are now scrambling to find ways to label, filter, and downgrade this content, but there is no consensus on the best method.
Some companies have experimented with disclosure labels that tell users when content was created with AI. Others have built watermarking tools to mark model outputs. Still, these approaches require cooperation from AI developers and can be easy to remove. Human reporting systems are imperfect, but they are one of the most direct ways for platforms to understand what their users actually want to see.
That could make the new button less about deleting a post immediately and more about changing the incentives that encourage people to publish meaningless content. If users know that their analytics dashboard may include a note saying their post looks like AI slop, they might think twice before firing off another generic motivational message. The hope is that these small signals, repeated at scale, will push platforms away from the race to the bottom and toward something closer to genuine human communication.
Source: Gizmodo News