AI-Tell Check

Score how AI-like a piece of writing is before you publish, and fix it

Problem

Someone answered questions with AI on Zhihu and immediately got a seven-day mute for suspected AI-assisted content, without knowing how the platform detected it. Creators also worry their genuinely human paragraphs get false-flagged, or want to keep AI help but sound like themselves.

Solution

A local-first detection and rewriting assistant: an explainable AI-likeness score based on repetition, connector density, paragraph structure and punctuation habits, sentence-level highlighting, and rewrites that preserve meaning while reducing the AI feel, exportable in multiple versions.

Users

Content-platform writers, self-media authors submitting work, ghostwriters and marketing copywriters

MVP

Paste text for an explainable score and feature list; sentence-level highlighting; one-click three rewrite levels; runs fully in-browser locally; before/after comparison and history

Revenue

Free tier plus a per-volume subscription; a one-time pro desktop edition

Why now

Platforms are using AI detection to ban accounts, creating urgent self-protection demand and a huge transparency gap; a small on-device model suffices without large-scale training