LinkedIn has always been that place where everyone tries to look more professional than they actually are, but lately the platform went beyond that and became one of the biggest examples of how AI-generated content can swallow an entire social network.
If you have ever scrolled through the feed and felt something was off — that the posts sounded too artificial, too generic, too polished — just know you were not alone. 👀
The situation reached such a critical point that the platform had to take real action, and not with a nice press release or an updated terms of service. The answer was a button.
A simple button called seems like AI slop, which users themselves can click to flag posts that look like they came straight out of a text generator. The announcement came from LinkedIn’s own chief product officer, Hari Srinivasan, who posted on the platform that fighting AI slop had become a top priority for the entire team.
The results in the first two weeks were impressive: more than 1 million clicks, and a 40% drop in views for content flagged as slop by the platform, compared to the period just a few weeks before.
According to Srinivasan, despite the progress, there is still a lot of work to be done to make sure LinkedIn remains a place where you can find real people and real perspectives. In his words, that remains an ongoing concern.
But to understand why this matters, you need to look at the bigger picture.
The so-called dead internet theory, which for a long time was treated as a niche conspiracy theory, is proving to be increasingly real. Recent data from Cloudflare shows that, for the first time in history, AI-generated traffic surpassed human traffic on the web, with bots accounting for 61.9% of searches, versus just 38.1% coming from actual humans.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
And LinkedIn, with its more than 1.3 billion registered users, was at the epicenter of this problem. 🤖
What is AI Slop and why LinkedIn became the perfect example of the problem
The term slop might sound too informal to describe a phenomenon with such serious consequences, but it captures exactly what happens when platforms are flooded with AI-generated content that has zero quality or authenticity standards. It is that post that starts with In an increasingly connected world…, that throws around words like synergies and disruptive left and right, and wraps up with a generic question trying to force engagement in the comments. If you use LinkedIn with any regularity, you have seen this type of post dozens of times without even consciously realizing something was wrong — but still feeling that vague sense of unease that lingers after reading something that says absolutely nothing useful.
The problem on LinkedIn grew at a rapid pace because the platform, unlike networks such as Instagram or TikTok, has always had a very particular relationship with text. It is a network built around professional narratives, career stories, achievements, lessons learned, and market positioning. This means that any language model trained on internet data can mimic the LinkedIn tone with alarming ease, since the style is predictable, the patterns are repetitive, and the vocabulary is largely restricted to a specific set of corporate expressions. When AI models learned to replicate that style, the platform became a natural target for anyone who wanted to gain visibility without investing real time or creativity into their content.
The numbers back up that perception. An analysis by the AI detection startup Pangram, published in July, pointed to LinkedIn as the most saturated platform for artificial content of them all. More than 40% of long-form posts were flagged as entirely AI-generated, and LinkedIn posts alone accounted for nearly two-thirds of all the content the startup identified as artificial across its analyses. That puts the platform well ahead of other text-based networks like Medium, X, and Substack.
What sets slop apart from content that is simply bad is precisely this quality of superficially polished emptiness. A poorly written post by a human still carries traces of humanity, of context, of lived experience, even if it is poorly articulated. AI-generated slop has perfect grammar, flawless structure, and zero real substance. It is the text version of a stock photo: technically correct, emotionally hollow. And when this type of content starts to dominate a feed, it does not just pollute the experience — it erodes the trust that users place in the platform as a whole.
Dead Internet Theory: when the conspiracy becomes reality
The dead internet theory emerged in the late 2010s on alternative forums and tech circles as a disturbing idea: what if most of what we see online was not actually created by real humans? What if bots, algorithms, and automated content already silently dominated the web, creating an illusion of human activity while the real internet — the one made of genuine conversations during the era of blogs and forums — was essentially dead? For a long time, this idea was dismissed as paranoid exaggeration, the kind of thing that circulates in communities that enjoy elaborate theories. But the latest data is making it increasingly hard to ignore the theory as mere speculation.
Rudy Yang, a corporate and retail fintech analyst at PitchBook, summed up the situation well when he noted that there are a lot of missing pieces of information, but a good chunk of the data points in the same direction: there is more bot activity out there, and agentic AI is driving a large portion of the browsing traffic we are seeing.
The Cloudflare numbers are the most revealing so far: 61.9% of searches were generated by bots and automated systems, surpassing for the first time the traffic produced by actual humans, who came in at 38.1%. This does not necessarily mean the internet is literally dead, but it does mean that the proportion of genuine interactions between people is shrinking relative to the total volume of digital activity.
And it is not just traffic. A study by the Pew Research Center analyzed 10,000 web pages collected in July 2026 and found that 10% of them showed significant signs of AI authorship, compared to about 2% five years earlier. Even more striking: evidence of AI authorship was present in more than a third of all pages published after ChatGPT arrived on the scene in late 2022. When you combine that data with the explosion of AI-generated content on platforms like LinkedIn, the picture that emerges is one of an ecosystem where more and more content is produced by machines to be consumed by other machines, with humans increasingly becoming spectators of a conversation they technically started but no longer control.
What makes this scenario especially relevant for LinkedIn is the fact that the platform has always positioned itself as a space for human professional connection. Unlike a search engine, which has always been explicitly a tool, LinkedIn sells the idea that you are connecting with real people, learning from real experiences, building relationships that can change your professional trajectory. When AI-generated slop starts to dominate that space, it is not just cluttering up any random feed. It is directly attacking the platform’s core value proposition, which is the authenticity of human connections.
The button that became a social thermometer
LinkedIn’s decision to create the seems like AI slop button is interesting for several reasons that go beyond the technical functionality itself. From a product design standpoint, it is a bold choice because it puts the power to classify and eventually penalize other users’ content directly into the hands of users — something platforms tend to avoid because of the potential for abuse. Any collaborative flagging system can be gamed by coordinated groups looking to suppress specific voices, and LinkedIn certainly knows that. The fact that they moved forward anyway signals that the slop problem was assessed internally as a bigger threat than the risks that come with an open flagging tool.
The numbers from the first two weeks confirm the bet paid off. More than 1 million clicks on a flagging button is a data point that says a lot about the level of accumulated user frustration with AI-generated content on the platform. It is not the kind of thing people do out of curiosity or at random — at least not at that scale. It is the kind of behavior that indicates there was pent-up demand for a way to express exactly that feeling of this was clearly written by a machine and I do not want to see it in my feed. The platform created a mechanism for a frustration that already existed, and people used it immediately and massively.
The 40% drop in views for posts flagged as slop shows that LinkedIn is in fact using those signals to adjust the algorithmic distribution of content, and not just collecting feedback for internal reports. This is relevant because it means the platform is essentially creating a feedback loop where user behavior directly feeds the algorithm’s decisions about what to promote or suppress. It is a model that comes closer to what platforms like Reddit have been doing for years with upvote and downvote systems, bringing a layer of collective curation to an environment that until now operated almost exclusively on raw engagement — meaning how many likes, comments, and shares a post received, regardless of its quality or authenticity.
Beyond the button, LinkedIn also announced it will expand profile and page verification, and that it plans to remove the enhance your post feature — the one that let users rewrite an entire post with AI assistance. In its place, the platform will offer a tool focused on reviewing and correcting text, without rewriting the content from scratch. It is a subtle but symbolic shift: instead of making it easy to auto-generate posts, LinkedIn is now encouraging users to write in their own words and use AI only as a support tool.
What changes for people who create content on LinkedIn
For content creators who use LinkedIn as part of their professional visibility strategy, this move by the platform represents a pretty significant change of rules. Over the past few years, especially after tools like ChatGPT and its competitors became mainstream, a considerable portion of the content published on the platform started being generated or heavily assisted by AI, with varying degrees of human editing and personalization. Some creators only used the tools to structure ideas that were genuinely their own. Others completely outsourced production and published the model output with almost no changes. It is this second group that now faces concrete algorithmic penalties, and the distinction between the two cases is not always easy to make from a technical standpoint.
The real challenge that comes out of this new landscape is that the line between AI-assisted content and AI-generated content is genuinely thin and subjective. A person who uses a language model to fix grammar and improve the clarity of a text they actually wrote is using AI in a very different way from someone who types a prompt and hits publish without changing a single comma. But the algorithm — and even more so the users clicking the flagging button — cannot always distinguish between these cases with precision. What will likely happen is that the market will develop new writing and presentation standards that signal authenticity in ways that language models still cannot easily replicate, such as references to very specific experiences, genuinely controversial opinions, or narrative structures that diverge from the most common patterns in model training data.
There is also an important strategic dimension here for anyone who thinks of LinkedIn as a business channel. If the algorithm is going to start penalizing slop and promoting content that users classify as authentic, that means the competitive advantage will shift from those who can publish the most to those who can publish the best — which, ironically, was exactly the original premise of the platform before the explosion of AI-generated content. Publishing frequency without quality is going to stop being a viable strategy, and those who invested over the past few years in building a genuine, recognizable voice may see that turn into a very real advantage in the coming months. 🎯
LinkedIn is not alone in this fight
It is worth pointing out that the tech companies that opened the door to AI are now dealing with the flood of artificial content they themselves allowed in the beginning. And LinkedIn is far from the only one playing catch-up. Spotify, for example, reported in July that it removed 75 million bulk uploads and duplicate songs over the past twelve months — a huge chunk of the roughly 100 million uploads estimated on its platform.
Substack, for its part, announced a partnership with Pangram to launch an AI detection feature, allowing users to scan posts and comments to estimate how much of that content was created with the help of artificial intelligence. And perhaps the most dramatic case is Digg, the old link-sharing site, which in March actually laid off employees and temporarily shut down its app precisely because it could not keep up with bot-generated spam.
Digg CEO Justin Mezzell shared in a statement that as soon as the app beta launched, they immediately noticed posts from SEO spammers exploiting the fact that Digg still carried relevant link authority on Google. Within hours, they got a firsthand taste of what they had only heard about through rumors: the internet today is populated, in significant part, by sophisticated AI agents and automated accounts. According to him, the team knew bots were part of the landscape, but had no idea of the scale, the sophistication, or the speed with which they would find them.
LinkedIn’s move is one of the first large-scale examples of a platform using the collective intelligence of its users to fight the impact of AI-generated content on its ecosystem, and the early results suggest the approach has real potential.
Whether it will be enough to reverse the larger trend of the dead internet theory remains to be seen. But as a signal that the major platforms are finally taking the slop problem seriously, it is hard to ignore what just happened. 👊
