41% of Long LinkedIn Posts Are Fully AI-Generated

A study published July 9, 2026 by AI detection company Pangram Labs found that 41% of long-form LinkedIn posts — those over 250 words — read as fully AI-generated. LinkedIn made up roughly a third of the 1,002,627 posts scanned across five platforms, but accounted for 62% of everything flagged as AI-written across the entire dataset.

How the Study Actually Worked

Pangram built the dataset through a Chrome browser extension used by people who opted in to anonymized data sharing, scanning what those users actually saw scrolling their feeds — LinkedIn, X, Reddit, Medium, and Substack — between late April and the end of June 2026. That approach measures real, encountered content rather than a curated public sample, which is part of why the findings carry weight: this is closer to what an average LinkedIn user's feed actually contains than a study built from scraped public posts alone.

The Reddit Contrast Is the Clearest Signal

Reddit replies came back 98.1% human-written in the same dataset — a striking contrast to LinkedIn's numbers from a platform scanned under identical methodology. That gap suggests the difference isn't really about AI capability or availability; it's about incentive structure. A LinkedIn post is written to be seen by a professional network and often tied to personal branding, while a Reddit reply is typically anonymous and conversational — circumstances that appear to correlate strongly with how much AI involvement ends up in the final text.

LinkedIn's share of AI-flagged content being nearly double its share of scanned content isn't just "some AI use" — it's evidence that AI-generated posts get produced and posted at a genuinely disproportionate rate on that specific platform, relative to how much content actually exists there. Something about how LinkedIn is used — the pressure to post consistently, the professional performance angle, the volume expected from personal-brand and thought-leadership content — appears to push AI adoption further than anywhere else measured.

Length Matters More Than Platform, With One Exception

Across the combined five-platform dataset, longer posts were consistently more likely to be flagged as AI-generated than shorter ones — about a quarter of all content over 250 words came back as fully AI-written, roughly double the 13.8% average across all content regardless of length. LinkedIn was the clear outlier within that pattern, sitting well above every other platform at the long-form end specifically, rather than simply following the general length-correlates-with-AI trend everyone else showed.

A Second Study Found an Even Higher Number

A separate July 2026 study by Originality.ai, using a different methodology — sampling 5,000 publicly available long-form LinkedIn posts rather than scanning opted-in users' actual feeds — classified 81.2% of that sample as likely AI-generated, roughly double Pangram's figure. The two studies aren't measuring quite the same thing: a public sample of posts (which may skew toward more visible, algorithm-favored, possibly more AI-optimized content) versus a feed-scan of what real users actually encountered scrolling. Both point in the same direction — LinkedIn ahead of other platforms — but the specific percentage depends heavily on which population of posts gets measured.

What Pangram's Own Team Said About the Numbers

Pangram's CEO described the findings as representing "a tax on readers' time," and the company has characterized its own percentages as a lower bound rather than a ceiling — meaning the true rate of AI involvement, including lightly-edited AI drafts that wouldn't register as "fully AI-generated," is likely higher than even these already substantial figures suggest.

What This Means for Anyone Reading or Writing on LinkedIn

For a reader, the practical takeaway is that roughly two in five long LinkedIn posts encountered while scrolling are, by this measurement, not written by the person whose name is attached to them — a meaningfully different reading experience than a platform where that number sits closer to Reddit's near-zero rate. For anyone writing on LinkedIn specifically, standing out from AI-generated content increasingly means demonstrating something a fully-automated post can't easily replicate: a specific detail, a genuine anecdote, a viewpoint that couldn't have been generated from a generic prompt about the same topic.

---

Pangram's numbers — 41% of long-form LinkedIn content fully AI-generated, LinkedIn responsible for 62% of all AI-flagged material despite being a third of what was scanned — describe a platform where AI-written content isn't a minority curiosity anymore; it's a substantial share of what fills the feed. The Reddit comparison in the same dataset makes clear this isn't inevitable everywhere text gets posted online — it's concentrated specifically where the incentives point toward frequent, polished, professional-facing content.