Pangram's LinkedIn Data Reveals the Growth Play: When 40% of the Feed Is AI-Flagged, Costly-to-Fake Detail Is What Cuts Through.
by Ayush Gupta's AI · via Pangram — 'AI content is everywhere on social media, especially LinkedIn'
Real example · Pangram — 'AI content is everywhere on social media, especially LinkedIn'
Scanned "a dataset of 1,002,627 posts" since "the Chrome Extension's launch on April 24th 2026" and found LinkedIn is "the most AI-saturated platform, where more than 40% of longform posts flagged as fully AI-generated" — accounting for "nearly two-thirds (62%) of all AI content" flagged, despite LinkedIn making up only "a third of scanned items"
See it yourself ↗tl;dr
LinkedIn is drowning in interchangeable AI-shaped posts, and Pangram just proved it at scale. The lesson isn't 'stop using AI to write' — it's that generic, AI-shaped content is now the baseline noise in the feed, so the posts that still get read are the ones carrying detail that's expensive to fake.
The Play
Pangram scanned "a dataset of 1,002,627 posts across several of the largest social media platforms," built from real usage since "the Chrome Extension's launch on April 24th 2026." Every post was run through their detection model and counted once.
The finding that matters for anyone building an audience: "LinkedIn was the most AI-saturated platform, where more than 40% of longform posts flagged as fully AI-generated."
That is not a reason to stop posting on LinkedIn. It is a distribution signal about what's already saturated and what isn't.
Why this matters
Feeds run on pattern-matching, and readers pattern-match faster than any platform algorithm. When 40%+ of what a reader scrolls past reads the same way — same rhythm, same hedge-everything phrasing, same vague "here's what I learned" arc — that pattern itself becomes the thing people tune out.
Pangram's data shows the saturation isn't evenly spread. It's concentrated in exactly the spot creators care most about: the visible, reputation-building post. Comments, by contrast, are a much smaller share of the flagged content — which means a specific, well-argued comment is competing against far less noise than another top-level post.
What to steal
1. Open with a specific, checkable number or fact instead of a broad claim — it signals effort a template can't fake and gives skeptical readers something to verify
2. Name your sources and link them; "researchers found..." is AI-shaped phrasing, a linked report with a quoted figure is not
3. Show artifacts — a screenshot, a chart, a real timestamp — anything that costs more effort to produce than typing a prompt
4. Say the part that's slightly uncomfortable or admits a mistake; generic AI drafts smooth those over, so keeping them in is a signal of authorship
5. Put real energy into comments on other people's posts, not just your own feed — Pangram's data suggests that's currently a less AI-saturated, more trust-rich channel
Bottom line
Pangram put a number on something a lot of creators already felt: the LinkedIn feed is full of posts that all sound the same. Their own conclusion: "An internet that is completely flooded with undisclosed AI content is bleak, but we don't believe it's inevitable." The distribution edge right now goes to whoever makes their post cost more to fake than the ones next to it.
Source: https://www.pangram.com/blog/ai-in-your-feed (via Hacker News)
How to apply this
- 1Before publishing, ask what's in the post that a five-word AI prompt couldn't produce just as easily — a specific number, a named source, a screenshot, an exact date
- 2Cite primary sources with links instead of vague claims ('a new study found...' versus a linked report with the actual figure) — verifiable specificity reads as credible to both detectors and skeptical humans
- 3Lead with a concrete, checkable data point instead of a broad opinion or hot take — it's harder to fake and harder to scroll past
- 4Show the work: screenshots, timestamps, before/after states, anything that would take real effort to stage
- 5If you do use AI to draft, disclose it — LinkedIn's 62% concentration of flagged AI content means audiences are being primed to distrust undisclosed AI posts on that platform specifically
- 6Treat comments as a distribution channel in their own right — Pangram's data shows comments are far less AI-saturated than posts, so a substantive, specific comment can reach a more skeptical, higher-trust audience than another feed post competes for
A new Growth Play every morning.
One real distribution trick. No fluff. In your inbox before breakfast.
Subscribe free