The AI Saturation Crisis on LinkedIn | Yatra Rawat

Yatra Rawat looks at why so many LinkedIn posts have started to read the same way, and points to a study that measured the problem directly: researchers sorted a million posts across five platforms into three bins, written by a person, written with help from a machine and edited, or fully generated, and LinkedIn came out worse than any other platform in the scan.

What the study found

The detector split each post into one of three categories: human-written, machine-assisted and edited, or fully AI-generated. On LinkedIn, 41% of long-form posts, meaning anything past 250 words, fell into the fully generated bin. That is the highest share of any platform in the study. For comparison, Medium came in around 31%, X between 25% and 29%, Substack at 10%, and Reddit as low as 4%. Averaged across all five platforms, roughly one in four posts scanned was fully AI-written. LinkedIn accounted for about a third of the million posts scanned but nearly 62% of every AI-generated post found across the whole study, meaning it is doing most of the machine writing by itself.

The missing middle

The more revealing number sits in the middle bin, the one for posts written by a machine and then edited by a person. On LinkedIn that bin measured about 4.3%, close to empty. People are not touching up AI drafts before posting them. They are handing the post over whole, unedited, straight from the tool to the feed. That pattern is different from what you would expect if people were using AI as a drafting aid the way it's often pitched. Instead the choice looks binary: write it yourself, or let the machine write and publish it as is.

The structural contradiction

LinkedIn ships a feature called "enhance" that helps write your post for you, and at the same time the platform is understood to downrank AI-detected content in its feed algorithm. Yatra calls this the structural joke: the same company supplies the tool that produces the flood and then penalizes the output that results. Detection itself is a probability rather than a confession, so no single post can be proven AI-written with certainty. But the comparison across platforms still holds: Substack and Reddit, which don't push the same generation tools as hard, sit at the bottom of the same scan. That gap points at something about LinkedIn specifically, not at AI writing tools in general.

What this means if you post there

If you're building a personal brand on LinkedIn, the medium is now suspected by default, since a plurality of long-form posts in your feed were likely generated wholesale. Since August 2nd, the article referenced in the video notes that disclosure requirements are already in effect for AI-assisted content on the platform. The practical upside Yatra draws out is that if you actually write your own posts, using specific numbers, names, and things that went right or wrong in your own experience, that specificity is now a differentiator rather than table stakes.

Key takeaways

• A one-million-post study found LinkedIn has the highest share of fully AI-generated long-form content of any platform measured, at 41%.
• LinkedIn produced about 62% of all the AI-generated posts found across the entire five-platform study.
• The middle category, AI-assisted posts that were then edited by a human, measured only 4.3% on LinkedIn, suggesting people mostly publish AI drafts unedited.
• LinkedIn simultaneously offers an AI post-writing feature and downranks AI-detected content in its algorithm.
• Posts that include specific, unrepeatable facts, a number, a name, a real detail, are the clearest signal that a human actually wrote them.

Who this is for

This is for anyone posting regularly on LinkedIn who wants to understand why so much of the feed reads the same way, and for people building a personal or professional brand there who want a concrete way to make their own writing stand out from AI-generated posts.