LinkedIn’s AI Problem: Study Finds Majority Of Posts May Be Machine-Written

Subscribe to our FREE Newsletter, or Telegram and WhatsApp channels for the latest stories and updates.

A recent Threads post by user poseidanai has been living rent-free in our heads. Why? The implications of the post are kinda scary – apparently, LinkedIn might be dead.

So the post refers to a study done by Pangram, a fairly well-known company that provides AI detection services. They have a Google Chrome extension where users can opt into sharing data with the company for R&D purposes, and what they found was kinda disturbing.

A Significant Chunk Of The Internet Is Just AI-Generated

According to Pangram, almost 14% of all internet content scanned via its users – including X/Twitter, Medium, Reddit, Substack and LinkedIn – are likely to be generated using AI. This is across more than a million data points.

Think about that. About 3 in 20 pieces of online written content involves some form of generative AI.

That number jumps to a shocking 1 in 4 when it comes to long-form content – meaning anything that’s over 250 words.

And as we alluded to earlier…

Nearly Two-Thirds Of Content On LinkedIn Is Made With AI

Across the board, LinkedIn scored the worst. Content from the platform only made up about 30% of all scanned items, but about 62% of it was flagged as touched by AI.

This might be because LinkedIn has a built-in AI writing assistant branded as “Enhance post”, like something you’d see if you’re using Grammarly. As a result, the platform has garnered a reputation for AI slop.

LinkedIn appears to want to rein things in a bit, but ironically, a post made by an executive at the company announcing that it would be detecting and downranking posts made by AI… was made by AI.

Reddit, on the other hand, seems to be the bastion of humanity, so to speak, because 98% of posts made there were supposedly written by people. Seems like when people want to be casual, anonymous and/or weird, they still prefer to get their hands dirty.

Of course, Pangram’s study isn’t perfect, since the data they have is limited to the platforms mentioned above, and you might argue that stuff generated by humans should be distinct from stuff generated by bots, but…

Are We Headed Towards The “Dead Internet Theory”?

If you’re not familiar, the Dead Internet Theory is an old internet conspiracy theory that suggested the web was entirely taken over by bots, so the interaction online was just bots talking to bots.

When the theory first made the rounds in the early 2020s, it was thought of more as a scary story, but nowadays, it might be the reality we’re looking at every day with the advent of Large Language Models, and far beyond just the platforms Pangram looked at.

In 2023, a security firm Imperva report found that nearly 50% of Internet traffic was automated, a 2% rise from 2022, which was partly linked to AI models scraping the web for training content.

This is following a prediction made back in 2022 by Timothy Shoup of the Copenhagen Institute for Futures Studies, who warned that 99% to 99.9% of all online content could be AI-generated between 2025 and 2030.

Moral of the story? Go to Reddit if you want genuine human interaction, we guess, or come comment on our social media posts. No bots, promise.

Share your thoughts with us via TRP’s Facebook, Twitter, Instagram, or Threads.

...