LinkedIn and X Flooded With AI-Generated Content, Research Reveals
Key Takeaways
- ▸41% of longform LinkedIn posts analyzed were fully AI-generated, with additional posts using AI assistance
- ▸Approximately 25% of X articles are fully AI-written, with another 23% AI-assisted
- ▸Longform content is significantly more likely to be AI-generated across all platforms, while shortform content remains mostly human-written
Summary
Research from Pangram, a company that detects AI-generated writing using a Chrome extension, reveals that a substantial portion of the written content users encounter on social media is AI-generated. According to data collected from approximately one million posts across LinkedIn, Medium, X, Reddit, and Substack over two months, 41 percent of longform LinkedIn posts are fully AI-written, and roughly one-third of longer posts on X contain AI-generated content.
The research is significant because it measures what users actually encounter while browsing, rather than counting AI content more broadly across the internet. The findings show that AI-generated content is not sequestered on obscure spam sites but actively poisons the feeds of everyday users on major platforms. Longform content (posts longer than 250 words) is far more likely to be AI-generated than shortform posts across all measured platforms. Notably, users appear willing to use AI to create professional content on identity-tied platforms like LinkedIn, suggesting they are comfortable publicly misrepresenting AI-generated text as their own work.
CEO Max Spero of Pangram characterized AI-generated content as "a tax on readers' time." The research highlights the growing problem of what some call 'AI slop'—voluminous low-quality content produced by generative AI systems that degrades the overall information quality of the internet used by humans.
- Top-level posts on LinkedIn and Reddit are far more likely to be AI-generated than comments on those posts
- Users are more willing to use AI in professional settings with real identity association than on anonymous or casual platforms
Editorial Opinion
The findings paint a troubling picture of AI-generated content not just polluting the wider internet, but actively degrading the quality of information users encounter on major professional and social platforms. The willingness of users to present AI-generated content under their own names on identity-tied platforms like LinkedIn is particularly concerning, as it erodes trust in professional networks and expertise. Until platforms implement stronger detection and disclosure requirements, or generative AI companies build in better safeguards against this type of misuse, users will bear an increasing cognitive burden in assessing authenticity.
