What happened
Analysis of over one million posts across five platforms (April–June 2026) found that 41 percent of long-form content on LinkedIn was flagged as fully AI-generated, while X/Twitter had close to half of long-form articles as AI-generated or AI-assisted. Substack reported around 10 percent, and Reddit replies were 98 percent human-written. Overall, one in four social media posts over 250 words is AI-generated.
Why it matters
LinkedIn made up only a third of all posts scanned but accounted for nearly two-thirds of all detected AI content, suggesting the platform is becoming a hub for machine-generated text. LinkedIn has already begun cracking down on AI-generated posts in response to the pressure.
What to watch
Pangram's detection model claims a false positive rate of 0.01 percent, though the company notes it may be better at identifying human-written content than AI-generated content, so the real AI rate could be even higher than reported.
Summaries like this, in your inbox every morning.
LinkedIn's dominance in AI-generated long-form content reflects a growing trend of automated posting on professional networks. The platform's design—which emphasizes thought leadership and long-form articles—appears to make it a natural target for AI content generation. Despite representing only a third of posts analyzed, LinkedIn accounted for nearly two-thirds of detected AI content, a disparity that underscores how concentrated the problem is on this single platform.
The detection methodology carries an important caveat: Pangram's model may be better at identifying human-written content than AI-generated content, meaning the real proportion of AI posts could be even higher than the 41 percent reported. Notably, Reddit replies showed the opposite pattern—98 percent human-written—while standalone Reddit posts contained AI text far more often, suggesting that the format and audience of a platform shape how much AI content appears. LinkedIn's recent crackdown indicates the platform recognizes this as a reputational and trust issue, though the scale of the problem suggests such efforts will need to be substantial to meaningfully shift the trend.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
