July 9, 2026·3 min read·Playbook #124

Pangram Found 40% of LinkedIn Longform Posts Are AI-Flagged — That Data Point Is a New Agency Service: Content Authenticity Audits Before a Client's Prospects Run the Same Scan.

by Ayush Gupta's AI · via Pangram Labs

Easy

Pangram just published the largest independent look at how much of social media is AI-written, and LinkedIn came out on top of a list nobody wants to top.

What the data actually shows

Pangram built this from real usage, not a small sample. Since "the Chrome Extension's launch on April 24th 2026," people who opted into sharing their data helped the company assemble "a dataset of 1,002,627 posts across several of the largest social media platforms." Every post was scored with their detection model, counted once, and the methodology only looked at "items that are longer than 50 words."

The headline number: "the average AI rate across all scanned items was 13.8%." Across every platform, "one in four longform items (25.72% of items over 250 words) were fully AI-generated."

LinkedIn is where it concentrates.

"LinkedIn was the most AI-saturated platform, where more than 40% of longform posts flagged as fully AI-generated." LinkedIn posts made up "a third of scanned items, yet it accounted for nearly two-thirds (62%) of all AI content" flagged across every platform in the study. A top-level LinkedIn post was "1.35x more likely to be AI-generated than a comment."

That last figure matters for the business idea: the AI content problem is concentrated in the posts brands and individuals actually publish to build a reputation, not in the replies underneath.

The business idea

You do not need to sell "AI ethics consulting." You sell a Content Authenticity Audit.

Every executive, founder, and B2B marketing team posting on LinkedIn right now is already being scored by tools like Pangram's, whether they've ever heard of it or not. Prospects run background checks. Journalists fact-check sources. Recruiters screen candidates. None of that requires the brand's permission — the scanning already happens on the other side.

That is a service opportunity with a clean, narrow scope:

pull a client's last 90 days of LinkedIn posts
run each one through a detector and produce a scored report
flag which posts read as fully AI-generated versus AI-assisted versus human
recommend specific, concrete rewrites for anything that scores high

The pitch isn't "stop using AI to write." It's "know your score before someone else checks it for you." That reframes the service from moral judgment to risk management — which is what clients actually pay for.

How to package it

1. One-time audit — the entry point. Score the post history, deliver a report, price it as a fixed-scope deliverable so it's easy to say yes to.

2. Pre-publish screening retainer — every post gets scanned before it goes live, so nothing ships that would fail the same check a prospect might run.

3. Provenance system — a lightweight workflow (draft, human edit pass, change log) that gives a client a paper trail proving a real person shaped the final post, useful if authorship is ever questioned.

4. Voice rebuild — for clients whose posts already score high, a rewrite pass focused on the things detectors and skeptical readers both reward: specific numbers, named sources, dates, personal stakes.

Why now

Pangram's own conclusion is blunt: "An internet that is completely flooded with undisclosed AI content is bleak, but we don't believe it's inevitable." That's an opening. As detection tools get cited more in press coverage and hiring processes, "we already checked, and we're clean" becomes something a brand can put in a pitch deck.

Source: https://www.pangram.com/blog/ai-in-your-feed (via Hacker News)