LinkedIn is flooded with AI written content, new report shows... but should we trust it

A new study from Pangram, the company behind a major AI text detection tool, claims that a surprisingly large portion of LinkedIn content is fully AI generated. According to their data, 41 percent of longform LinkedIn posts and 30 percent of short posts are flagged as entirely AI written, making LinkedIn one of the most AI saturated platforms. Medium follows closely with 31 percent of longform content, while X shows 29 percent for longform posts and only 9 percent for regular short posts. Reddit and Substack show lower but still notable levels of AI generated writing.

Editor’s Note There is a fundamental flaw in how studies like this interpret AI generated content. Tools such as Pangram, Turnitin, and Copyleaks often label a post as “100% AI” simply because the author used a language model to translate or refine text written in a non-native language. The ideas, arguments, and reasoning remain entirely human, yet the polished English produced during translation triggers statistical patterns that detectors misread as automated writing.

A large portion of that supposed 41 percent of AI generated LinkedIn content is almost certainly written by real professionals from Eastern Europe, India, Latin America, and other regions where English is not the first language. These are genuine voices using technology to express their own thoughts clearly and professionally, not bots flooding the platform with synthetic posts. An algorithm can analyze sentence structure, but it cannot determine who conceived the idea behind the words. By treating linguistic refinement as artificial authorship, detectors end up mislabeling authentic human work as machine made content.

Read the full story on Gizmodo →