Does this post sound like AI?
Paste a draft. You get a human score, every pattern that gives it away as AI writing, and what to write instead. Nothing leaves your browser.
Runs in your browser. Nothing you type leaves this page. No signup.
- -14 Stock AI phrasing: “here's the truth”, “the result?”. Say it the way you would out loud.
- -10 Engagement bait: “agree?”. End on a specific question or nothing.
- -8 Model vocabulary: “seamless”, “transformative”. Use the plain word.
- -8 “It's not X. It's Y.” framing. Make the claim directly; the contrast is a template readers spot.
- -7 1 “The result? …” reveal. State the result in the sentence that earns it.
- -6 A moral at the end (“Remember: …”). Stop when the point is made.
Sentences average 3.9 words (spread 1.8). 12.9 contractions per 100 words. Under 75 reads as AI to most people who scroll past.
GrowedditX writes from your own stories and numbers, and rewrites any draft that fails this check before you see it.
Rewrite it so it sounds like meHow the detector works
- It looks for the phrases, words and shapes people now recognise as AI writing: the reveal question (The result?), the not this but that frame, lists of three, one sentence per line, a moral at the end.
- It measures rhythm. People mix a three word sentence with a twenty word one; models keep every sentence about the same length.
- Each finding takes points off 100 and comes with the fix. Under 75 reads as AI to most people who scroll past it.
Why sounding human matters for reach now
X labels accounts its classifier judges as AI slop and drops their posts from the For You feed for seven days. The rule is in X's open source ranking code, published at github.com/xai-org/x-algorithm.
LinkedIn added a way for members to report a post as AI slop in July 2026, and it hides comments it detects as automated.
The fix is not a better prompt. It is material only you have, your own rhythm, and none of the stock moves this page flags.
Sources
Checked on 30 September 2026. Platforms change their limits; check the source if you are close to one.
Questions
Does this detect whether a model wrote my post?
No tool can prove that. This checks for the patterns readers and platforms treat as AI writing, which is what decides how your post is received. A human can trip them too.
Does my post leave my browser?
No. The check runs in this page and nothing is sent anywhere.
What score should I aim for?
75 or more. Above 90 is rare for a first draft. Fix the biggest finding first; most posts jump 20 points from one rewrite.
Your next post, in your voice, on both networks.
7 days free. No card. Your drafts are kept.