Sounds Like AI? — AI Writing Authenticity Checker

Project Title: Sounds Like AI? — AI Writing Authenticity Checker The problem: Most professionals now write with AI assistance, but that comes with a side effect — posts that read as generic, over-polished, or robotic. There was no quick way to catch this before publishing.

What I built: An AI agent that analyzes any piece of writing (starting with LinkedIn drafts) and scores how "human" it reads. It flags specific patterns — generic openers, corporate jargon, overused transitions, cliché phrasing — and explains why each one reads as AI-generated. Then it generates a rewritten version with more personal voice, and saves every analysis so users can track improvement over time.

How it works: The app sends the input text to Llama 3.3 (via Groq) with a structured prompt that returns a JSON response — score, flagged phrases with reasoning, and suggested rewrites. This gets parsed and rendered into a clean, editorial-style UI with a live score gauge, inline highlighting, and a persistent history log backed by a database.

Why it's different: Most "AI detectors" just give a binary flag. This one explains its reasoning like a writing coach and fixes the problem in the same flow — built for someone to actually act on, not just read. Tech stack tags: Groq, Llama 3.3, AI Agents, Python/Backend, Database, UI/UX Design, Replit