CliNe

A Chrome extension that flags clickbait headlines while you browse. LSTM model, small Flask API, score in a popup.

Why

I kept clicking headlines that oversold the article. I wanted something in the browser that could warn me before I opened the tab.

CliNe Chrome extension banner

How it works

An LSTM trained on labeled headlines with GloVe embeddings. About 93% on trending titles from live news feeds.1 A Flask API on Render returns the score; the extension shows it in a popup.

You can also report false positives and false negatives from the extension, so the model can improve after launch.

Figma canvas with CliNe logo, popup, and UI shots while designing the extension
Figma canvas while we designed it: logo, popup, shots.

In the browser

On a news site it scans the headlines on the page, marks the ones that look like bait, and shows the score in the popup if you want more detail.

CliNe Chrome extension highlighting headlines on a news page
Live on a news page.
CliNe project homepage
Project site.

The API

The extension calls a small Flask service. Source is on GitHub. POST a headline to /predict:

{
  "text": "You will not believe what happened next"
}
Request
{
  "label": "clickbait"
}
Response

The label is clickbait or not clickbait. Extension code is here.

Footnotes

  1. Scored on headlines from live news feeds, not a cleaned-up benchmark set. ↩