Skip to main content

LinkedIn flooded with AI spam: 41% of long posts are machine-written

Social media is increasingly becoming a playground for AI-generated content, and a recent study suggests that LinkedIn is leading the pack in the wrong direction. According to data analysis company Pangram, a staggering 41% of long-form posts on the professional networking platform are now written by artificial intelligence.

The study, conducted between April and June 2026, analyzed over 1 million posts across five major social media platforms using Pangram's Chrome extension. The results paint a clear picture: AI is everywhere, but it's especially rampant on LinkedIn. Despite accounting for only one-third of all scanned samples, the platform contributed nearly two-thirds of the detected AI content. Even short posts (50-250 words) on LinkedIn saw a 30% AI generation rate—the highest among all platforms surveyed.

Image

Medium, the blogging platform, came in second with 28% of short-form and 31% of long-form content flagged as AI-generated. X (formerly Twitter) also showed significant AI involvement in longer posts, with nearly half of its long articles either fully generated (29%) or AI-assisted. However, short-form content on X remained relatively human, with only 9% AI involvement.

On the flip side, some platforms have managed to keep AI at bay. Substack boasted the lowest proportion of AI-generated long-form content at around 10%, while Reddit showed a unique pattern: although individual posts had a relatively high frequency of AI text (13% for long-form, 3% for short-form), a whopping 98% of user replies were still written by real people. This suggests that while bots may be posting, the community is still actively engaging in human conversation.

The Pangram3 detection model used in the study claims a high accuracy in identifying human-written content, with an error rate of only 0.01%. This means the actual proportion of AI-generated content on social media could be even higher than reported. As the problem grows, LinkedIn itself has started taking measures to combat AI spam, but the battle is far from over.

Key Points

  • LinkedIn is the worst offender: 41% of long posts and 30% of short posts are AI-generated.
  • Medium and X also affected: AI content is prevalent on these platforms, especially in longer posts.
  • Substack and Reddit stand out: Substack has the lowest AI content, while Reddit's replies remain largely human.
  • Detection model is accurate: Pangram3 has a 0.01% error rate, suggesting the problem may be even bigger.
  • Platforms are responding: LinkedIn has begun cracking down on AI-generated posts.