LLM Cliché Highlighter by Haider Alwasiti

Ever read something online and think "did a chatbot write this?" This add-on shows you which bits gave it away.

HOW IT WORKS

Click the button in your toolbar. The article lights up: every phrase that matches a known AI writing habit turns yellow. Tap or hover one and it tells you which habit it spotted. The number on the button shows how many it found. Click the button again and the page goes back to exactly how it was.

WHAT IT LOOKS FOR

38 patterns in all — the phrases AI writing reaches for again and again:

• "No sign-ups, no downloads, no hassle" — the three-part list
• "It's not just an office, but a small museum"
• "I won't pretend it was smooth" — the humble admission
• "It is important to note that..."
• Words like "delve", "a testament to", "ever-evolving landscape"

A highlight means "worth a second look", not "a robot wrote this". Plenty of people write this way, and plenty of AI text won't trigger anything at all. It is a reading aid, not a detector.

YOUR PRIVACY

It cannot spy on you, and that is built in rather than promised:

• It has no access to any page until you click the button — and then only to that one tab.
• It has no permission to use the internet, so it cannot send anything anywhere.
• It saves nothing. Close or reload the page and the highlights are gone.
• No accounts, no analytics, no tracking.

GOOD TO KNOW

• It reads the main article. Menus, sidebars and footers are left alone.
• It looks at the page as it is when you click, so on sites that load more as you scroll, click once the text has appeared.
• It does not look inside embedded frames.

CREDITS

Free and open source, under the Apache License 2.0.

The pattern-matching engine is Simon Willison's llm-cliche-highlighter. This add-on is a fork of Bruno Renié's llm-cliche-highlighter-extension, which turned that engine into a browser add-on. Both are credited in the licence file bundled with the add-on.

Source code: https://github.com/hwasiti/llm-cliche-highlighter-extension