Reduce detectable AI-style evidence in your writing
Targets the passages that actually carry evidence, on your device, with no server involved, then shows you what a detector still measures in the result. It cannot guarantee defeating a model vendor's undisclosed watermark; see what it does and doesn't claim below.
Paste, drop a file, or upload .txt, .md, .rtf or .csv. Everything is read and processed on your device.
Your text is processed in this browser tab and never uploaded. Here is how that works.
How it works, and its limits
Can this guarantee my writing won't be flagged?
No. It reduces detectable evidence, both statistical watermark signal where structurally possible and human-perceptible AI tells like em dashes and stock phrasing. It cannot guarantee defeating a model vendor's undisclosed watermark, because nobody outside that vendor holds the key it was applied with.
Where does my text go?
Nowhere. Rewriting runs entirely on your device, on every tier, on every surface (browser, MCP, local package/CLI). There is no hosted mode for this feature, unlike checking.
Why does it sometimes leave a passage unchanged?
Every candidate rewrite has to preserve the numbers, negations and named entities in the original, and stay above a similarity floor for the strength you chose. A passage with no candidate that clears both is left completely unchanged rather than replaced with something unsafe.
What is the difference between the Standard and Pro engines?
Standard is deterministic substitution against the core AI-tell library: instant, no download, and included on every plan. Pro is a language model that runs in your browser, with the extended AI-tell library and more candidate rewrites per passage. It comes with a Pro subscription, which also removes the free plan's weekly budget of 20,000 rewriting tokens every 7 days. Checking is unlimited on both.