LinkedIn is reducing the spread of low-effort AI-generated content that sounds polished but provides little original perspective. The platform is building detection systems to distinguish posts with genuine expertise from repetitive, generic, and empty posts. Early tests claim 94% detection accuracy for generic content. Instead of removing flagged posts, LinkedIn will suppress them from recommendations so they remain visible to direct connections but do not spread widely. The effort targets engagement bait, recycled “thought leadership” without originality, and formulaic AI patterns such as “it’s not X, it’s Y.” The crackdown also applies to bot-generated generic AI comments and automation tools that generate content at scale, while allowing AI-assisted content when it includes original ideas.
"LinkedIn says it is building detection systems trained to distinguish between posts that add genuine perspective and posts that feel repetitive, generic, and empty. In early tests, LinkedIn says its system correctly flagged generic content 94 per cent of the time. Flagged posts will not be removed. Instead, they will be suppressed from recommendations, meaning they will still be visible to a poster's direct connections but will no longer spread across the wider feed."
"The targets are specific. LinkedIn is going after outright engagement bait, recycled “thought leadership” that lacks originality, and posts with obvious AI construction patterns. The company singled out the “it’s not X, it’s Y” format as one example of the kind of formulaic AI content it plans to demote."
"The crackdown extends to comments too. LinkedIn will target bot-generated and generic AI comments that add nothing to a conversation, the kind that read like a ChatGPT summary of the post they are replying to. The platform is also going after automation tools that generate AI content at scale."
"There is, however, a deliberate line being drawn. LinkedIn says AI-assisted content is still welcome, provided it contains original ideas or en"
Read at TNW | Apps
Unable to calculate read time
Collection
[
|
...
]