LinkedIn shipped a feature that lets any stranger flag your post as AI slop. If you post content built with AI systems, that should probably scare you. It doesn’t scare me, and here’s the part most people are missing: the button isn’t detecting AI. It’s detecting emptiness.
Nobody has built a reliable AI text detector. OpenAI tried and pulled their own classifier because it was wrong too often to be useful — they said it themselves. LinkedIn’s version isn’t running some magic model that sniffs out GPT syntax. It’s a report button. It works the way every report button works: people flag things that feel off, and a human or a weak classifier makes a judgment call afterward.
What Actually Gets Flagged
Watch what people report as “AI slop” and it’s never actually about the tool. It’s about the writing having no fingerprint. No specifics. No stakes. No version of a real event that only happened to one person. Just smooth, confident sentences that could sit on any founder’s profile, any week, forever.
That’s not an AI problem. That’s a ghostwriting problem, a lazy-prompting problem, a nobody-edited-this problem. You can write voiceless slop by hand. You can run a post through five AI passes and end up with something sharper than anything you’d have written alone. The tool was never the variable. The editor was.
The 3-Second Test
Before anything goes live under my name, it runs through three questions. Takes less time than posting itself.
- Would a stranger know this happened to ME, not “a founder”? If you could swap your name for a placeholder and the post reads identically, there’s no fingerprint in it.
- Could I have written this exact line six months ago, word for word? If yes, you haven’t learned anything new since then, or you’re recycling a take instead of reporting from the ground.
- Does it sound like I’m talking, or like I’m presenting? Read it out loud. Presenting has topic sentences and clean transitions. Talking has fragments, a correction mid-thought, a specific number you actually remember.
Fail one of those and it doesn’t matter who typed it. It still reads like it belongs to nobody, and that’s the thing that gets reported — by a stranger, by the algorithm’s suppression, or just by silence in the comments.
The 4 Rules I Run Every Post Through
The test above is the fast filter. These four rules are what I actually build the post against before it ships.
- Rule 1 — One real number, not a range. “Grew engagement” is nothing. “[X] comments on a post that usually gets 40” is a fingerprint. If you don’t have the number yet, don’t publish the claim yet.
- Rule 2 — One decision you’d defend in a comment fight. Safe, agreeable takes are the easiest thing for a reader to skim past and the easiest thing for a stranger to write off as filler. Pick a side. Say the thing that could get pushback.
- Rule 3 — Cut the opening sentence that sounds like a LinkedIn post. Anything starting with “In today’s fast-paced world” or “Here’s why X matters” gets deleted. Start mid-thought, the way you’d start telling a friend at a bar.
- Rule 4 — Read it out loud before it goes out. If you stumble, so will the algorithm’s read on it and so will the reader. Reels script writers already know this. Text posts need the same discipline.
None of these rules mention AI. That’s the point. They’re voice rules. Run a fully AI-drafted post through them and it survives. Run a fully human-written post through them and skip a rule, and it reads exactly like the slop everyone’s suddenly panicked about.
Why This Matters More Than the Button Itself
The report button is a distraction from the actual shift happening on the platform. Reach isn’t being rationed based on how a post was drafted — it’s being rationed based on completion rate, dwell time, and comment quality, the same signals that have driven the feed for years. If you want the full mechanics of what’s actually moving reach right now, the LinkedIn algorithm audit breaks down what the platform rewards independent of any AI panic.
Format matters here too. A voiceless carousel gets skimmed and skipped just as fast as a voiceless text post — the 51-post breakdown on carousel vs. text shows where each format actually earns attention versus where it just looks busy. And if you’re chasing reach numbers without a plan for what they convert to, read the 200-person math piece before you optimize any further for impressions.
The Real Filter
Strangers can’t detect AI. What they can detect, in about the same three seconds you spend deciding whether to keep scrolling, is whether a real person is on the other end of the sentence. LinkedIn’s engineering team has said as much about how they think about feed quality — their own engineering blog is built around signals of genuine engagement, not authorship forensics.
So build with whatever tools get the draft done faster. Then run it through the test and the four rules before it ever goes live. The button was never coming for the AI. It was coming for the posts that never had anyone behind them in the first place.