An analysis of 1M+ social media posts from April 24 to June 30: ~25% of longform posts with 250+ words were fully AI-generated; on LinkedIn, the figure was 41%
A first look at analytics from our Pangram Chrome extension — Two months ago, we launched our Chrome extension to help combat the rising slop problem on social media.
Pangram LabsMax Spero
Context & Ripple Effects
Pangram’s latest sample extends its monitoring of AI-authored text from social feeds into other high-stakes written settings: its prior analysis found signs of extensive AI use in ICLR 2026 peer reviews. The firm has also launched a Chrome extension aimed at identifying low-quality AI-generated social content.
LinkedIn is especially consequential because it is simultaneously building creator-marketplace tools and courting influencers to increase engagement. Its elevated share in Pangram’s sample puts content authenticity alongside the platform’s push to make professional posting more valuable to marketers and audiences.
First-order effects
Pangram’s findings give users and publishers of long-form social posts a concrete reason to scrutinize provenance, particularly on LinkedIn, while creating an immediate use case for its detection extension.
For LinkedIn, the analysis raises a quality and trust issue around the feed just as creators and marketers are being given more ways to discover and evaluate one another.
Second-order effects
Creator-marketplace participants may place greater weight on signals beyond posting volume or polished copy—such as demonstrable expertise and audience trust—if AI-written posts make surface-level content less informative.
Detection products and platform moderation systems gain a clearer commercial and product rationale, while creators who rely on original analysis face pressure to distinguish their work from automated output.
Third-order effects
If AI-authored professional content remains common, social platforms may shift from maximizing content supply toward proving provenance, authenticity, and usefulness; how reliably detectors can support those judgments remains a limiting factor.
The pattern also points to AI-text governance becoming a cross-domain issue, spanning public social feeds and expert-review workflows rather than a problem confined to low-value content.
The trend: Generative AI is turning text abundance into a trust-and-verification problem for platforms and institutions that depend on written contributions as signals of expertise.
How much of what you're reading on social media is AI-written? LinkedIn and Twitter/X come out worst in long-form posts, with almost half AI-generated or AI-assisted, according to a new analysis. Based on posts of more than 250 words checked by the AI-writing detector @pangram, […
As much as 40 percent of longform posts on LinkedIn and X users actually see are AI-generated, according to new research Particularly notable because we know these platforms are flooded with slop, but this studied what the algorithm actually shows people https://www.404media.co/.…
Two months ago, we launched a Chrome extension that flags AI content in your feed. We included an option for users to share anonymized detection data with us. In that sample, one in every four longform posts was fully AI-generated. 🧵 [image]
If you're seeing long posts on LinkedIn, chances are pretty good that artificial intelligence wrote some or all of the post, according to new research highlighted by the crack team at 404 Media. …
“This isn't something that had really been studied before—how much AI content people are actually seeing,” Max Spero, the CEO of Pangram, told me in a phone interview. “AI content is a tax on readers' time.” — Read now: www.404media.co/linkedin-and...
We recently submitted a proposal titled “AI Everywhere All at Once” to study downstream implications of this AI onslaught on online communities. But we emphasized that claims of detection will be exaggerated; and already shown that AI benefits will be overblown. — www.pangram.…
🚨 The rising slop problem! 🗑️ — AI Content Is Everywhere on Social Media www.pangram.com/blog/ai-in-y... “...more than 40% of longform posts flagged as fully AI-generated.” — “Nearly half of X/Twitter articles contained AI writing.”