Back in my second year of college, I started noticing something strange online. Every website I visited felt the same. The writing style, comments, and “top 10” articles all looked repetitive. Even social media replies seemed robotic. One day, I saw two different accounts post the exact same paragraph within seconds, and that made me curious.
At first, it was just a joke between my friends. We called it the “Dead Internet Theory” project. But the more I explored, the more I realized how much AI-generated content and bot activity were flooding the internet. I wanted to know if there was a way to actually measure it.
So I started building a small Python script that analyzed text patterns from websites. Initially, it only checked repetitive words and sentence structures. The results were terrible at first because it kept labeling normal blog posts as bots. But that failure pushed me deeper into machine learning and NLP.
I spent nights learning about:
Perplexity analysis
Stylometry
Transformer models
Bot behavior patterns
Web scraping
After a few weeks, I connected the script to a simple dashboard. The first successful test happened when the system detected a fake review website with hundreds of duplicated AI-written comments. That moment made me realize the idea had real potential.
Then I expanded the project into a full platform called Dead Internet Detector.
The system now:
Crawls websites
Extracts text and metadata
Detects AI-generated writing patterns
Analyzes suspicious engagement behavior
Generates a “Human Authenticity Score”
One of the biggest challenges was reducing false positives because I did not want real human writers to be unfairly flagged. So I trained the model using both AI-generated and authentic human-written datasets. Over time, the detection accuracy improved significantly.
What started as a late-night curiosity project slowly became one of the most ambitious things I had ever built. The project combined AI, cybersecurity, NLP, and full-stack development into a single platform focused on solving a modern internet problem.
Today, the idea is evolving into something bigger than a college project. I want it to become a tool that helps people identify authentic content in a world increasingly dominated by automation and synthetic media.
Log in or sign up for Devpost to join the conversation.