Paste a draft. The checker flags the patterns readers have learned to associate with AI writing — giveaway vocabulary, formulaic constructions, hedging filler, em dash overload, uniform rhythm — and highlights every match.
Runs in your browser. Nothing you paste is uploaded, stored or logged.Works best on 100+ words. Not an AI detector — it flags specific patterns, shows you where they are, and leaves the judgement to you.
Deleting "delve" doesn't make content sound like you — it makes it sound like AI minus one word. The tells show up because the model wrote in its default voice. The fix is giving it yours: how to make AI content sound human covers the rewrite tactics, and the 12-point AI content audit is the deeper manual version of this checker.
If your posts keep triggering the same flags, the upstream fix is a voice prompt calibrated to your actual writing. Build a starter one free with the Voice Prompt Generator, or read why LinkedIn appears to downrank obvious-AI patterns.
The Voice Build calibrates an AI voice system against your existing writing — the tells never appear because the output starts in your voice, not the model's. 65 ready-to-publish pieces included.
See The Voice BuildNo. It doesn't guess whether AI wrote your text. It flags specific patterns readers associate with AI writing and shows you exactly where they are. Humans use these patterns too — the judgement stays with you.
No. The scan runs entirely in your browser. Nothing is sent to a server, stored or logged.
Four categories: giveaway vocabulary (delve, tapestry, elevate, robust), formulaic constructions ("it's not just X, it's Y"), hedging filler ("can be a great way to"), and rhythm problems (uniform sentence length, em dashes in every paragraph).
Not necessarily. One em dash is style; nine in 300 words is a signature. Read each flag and ask whether it's genuinely how you write or a habit picked up from reading AI output.
LinkedIn appears to downrank obvious-AI patterns rather than AI content per se. Removing the pattern signature helps. The stronger fix is producing from a calibrated voice prompt so the tells never appear. More on posting AI content without suppression.